stealthmark / index.html
stealth-model's picture
Map guide lines on hover and selection, Inspect links to model pages, game result links fixed
e213db2 verified
Raw History Blame Contribute Delete
147 kB
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>StealthMark · Independent AI benchmarks by Stealth Models</title>
<meta name="description" content="Physical and spatial reasoning, planning across many steps, and turning ideas into precise visuals. Every score on this page comes with the result behind it.">
<meta name="color-scheme" content="dark">
<link rel="icon" href="/_shared/favicons/stealthmodels.svg" type="image/svg+xml">
<link rel="preload" href="/assets/fonts/dm-sans-latin.woff2" as="font" type="font/woff2" crossorigin>
<link rel="stylesheet" href="/assets/space.css">
<script>document.documentElement.classList.add('js')</script>
<script type="module" src="/assets/space.js"></script>
</head>
<body>
<a class="skip" href="#scores">Skip to results</a>
<main>
<section class="hero" aria-labelledby="title">
<header class="masthead"><div class="wrap"><p class="kicker"><a class="plain" href="https://stealthmodels.com/?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">Stealth Models</a> · Independent AI benchmarks</p></div></header>
<div class="wrap">
<div class="prelude">
<div class="prelude-main">
<h1 id="title"><span class="wordmark">StealthMark</span> <span class="version" data-version>1.8.8</span></h1>
<p class="hero-line">Original tests no model could study for.</p>
</div>
<div class="prelude-side">
<p class="hero-lead">Physical and spatial reasoning, planning across many steps, and turning ideas into precise visuals. Every score on this page comes with the result behind it.</p>
<p class="snapshot" data-snapshot>Results as of 28 October 2026</p>
<p class="hero-actions"><a class="btn btn-gold" href="#scores">See the scores</a><a class="btn btn-outline" href="#evidence">See the evidence</a></p>
</div>
</div>
<div class="scene">
<div class="map-shell" data-map-shell>
<stealthmark-map compare="all" x-axis="general" y-axis="visual" range="full" routers="all" type="all" presentation="embed" theme="dark" aria-label="StealthMark 1.8.8 map: General Intelligence against Visual Intelligence"><p class="map-note" role="status">Loading map</p></stealthmark-map>
<noscript><p class="map-note">Use the results table below.</p></noscript>
</div>
<aside class="rail" aria-labelledby="rail-title"><h2 class="rail-title" id="rail-title">Current leaders</h2><ol class="rail-list" data-leaders><li class="rail-row rail-row--overall" data-leader="overall"><p class="rail-cat">Overall</p><p class="rail-score"><span class="sr-only">Benchmark score </span>74.7</p><div class="rail-ids"><div class="rail-id"><a class="lockup lockup--l1" href="https://stealthmodels.com/benchmarks/models/gpt-6-astra?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="openai/gpt-6-astra"><span class="mount"><img src="/benchmarks/brands/openai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="GPT-6 Astra">GPT-6 Astra</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div></div></li><li class="rail-row rail-row--open" data-leader="open"><p class="rail-cat">Open weights</p><p class="rail-score"><span class="sr-only">Benchmark score </span>59.2</p><div class="rail-ids"><div class="rail-id"><a class="lockup lockup--l2" href="https://stealthmodels.com/benchmarks/models/deepseek-v4.1-flash-1017?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="deepseek/deepseek-v4.1-flash-1017"><span class="mount"><img src="/benchmarks/brands/deepseek.svg" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="DeepSeek-V4.1-Flash-1017">DeepSeek-V4.1-Flash-1017</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div></div><p class="rail-links"><a class="sec" href="https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · DeepSeek-V4.1-Flash-1017, Official weights on Hugging Face</span></a></p></li><li class="rail-row rail-row--local" data-leader="local"><p class="rail-cat">Local</p><p class="rail-score"><span class="sr-only">Benchmark score </span>51.7</p><div class="rail-ids"><div class="rail-id"><a class="lockup lockup--l2" href="https://stealthmodels.com/benchmarks/models/qwen3.8-27b?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="qwen/qwen3.8-27B"><span class="mount"><img src="/benchmarks/brands/qwen.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Qwen3.8-27B">Qwen3.8-27B</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div></div><p class="rail-links"><a class="sec" href="https://huggingface.co/Qwen/Qwen3.8-27B" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · Qwen3.8-27B, Official weights on Hugging Face</span></a></p></li></ol></aside>
</div>
</div>
</section>
<section class="compare" id="compare" aria-labelledby="compare-title">
<div class="wrap">
<h2 id="compare-title">Compare any two models</h2>
<p class="section-lead">Every benchmark side by side, with the results behind each score.</p>
<div class="cmp" data-cmp data-a="openai/gpt-6-astra" data-b="deepseek/deepseek-v4.1-flash-1017">
<div class="cmp-grid">
<div class="cmp-cell cmp-id cmp-id--a" data-cell="id-a"><p class="cmp-label">First model</p><div class="cmp-idbox" data-side="a"><div class="cmp-lock"><a class="lockup lockup--xl" href="https://stealthmodels.com/benchmarks/models/gpt-6-astra?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="openai/gpt-6-astra"><span class="mount"><img src="/benchmarks/brands/openai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="GPT-6 Astra">GPT-6 Astra</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><button type="button" class="change" data-change>Change model</button><p class="cmp-status"></p><p class="cmp-sec"></p></div></div>
<div class="cmp-cell cmp-id cmp-id--b" data-cell="id-b"><p class="cmp-label">Second model</p><div class="cmp-idbox" data-side="b"><div class="cmp-lock"><a class="lockup lockup--xl" href="https://stealthmodels.com/benchmarks/models/deepseek-v4.1-flash-1017?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="deepseek/deepseek-v4.1-flash-1017"><span class="mount"><img src="/benchmarks/brands/deepseek.svg" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="DeepSeek-V4.1-Flash-1017">DeepSeek-V4.1-Flash-1017</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><button type="button" class="change" data-change>Change model</button><p class="cmp-status"></p><p class="cmp-sec"><a class="sec" href="https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · DeepSeek-V4.1-Flash-1017, Official weights on Hugging Face</span></a></p></div></div>
<div class="cmp-cell cmp-lab cmp-lab--a" data-cell="lab-a">First model</div>
<div class="cmp-cell cmp-lab cmp-lab--b" data-cell="lab-b">Second model</div>
<div class="cmp-cell cmp-n cmp-n--score cmp-n--a" data-cell="score-a">74.7</div><div class="cmp-cell cmp-m" data-cell="score-m">Benchmark score</div><div class="cmp-cell cmp-n cmp-n--score cmp-n--b" data-cell="score-b">59.2</div>
<div class="cmp-cell cmp-n cmp-n--gi cmp-n--a" data-cell="gi-a">73.4</div><div class="cmp-cell cmp-m cmp-m--gi" data-cell="gi-m">General Intelligence</div><div class="cmp-cell cmp-n cmp-n--gi cmp-n--b" data-cell="gi-b">56.8</div>
<div class="cmp-cell cmp-n cmp-n--vi cmp-n--a" data-cell="vi-a">77.0</div><div class="cmp-cell cmp-m cmp-m--vi" data-cell="vi-m">Visual Intelligence</div><div class="cmp-cell cmp-n cmp-n--vi cmp-n--b" data-cell="vi-b">63.5</div>
</div>
<p class="cmp-go-wrap"><a class="btn btn-gold cmp-go" id="cmp-go" href="https://stealthmodels.com/compare/gpt-6-astra-vs-deepseek-v4.1-flash-1017?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">Compare on stealthmodels.com</a></p>
<p class="cmp-note" data-cmp-note hidden role="status">Choose a different second model.</p>
</div>
</div>
</section>
<section class="scores" id="scores" aria-labelledby="scores-title">
<div class="wrap">
<h2 id="scores-title">Scores</h2>
<details class="means"><summary>What the scores mean</summary><dl class="definitions"><div><dt>Benchmark score</dt><dd>How well a model&#39;s abilities carry over to original, demanding tasks in reasoning, planning and visual work, as one score from 0 to 100. A high score means its skills hold up beyond familiar benchmark questions.</dd></div><div><dt>General Intelligence</dt><dd>The ability to reason, plan, solve problems and understand how things relate and behave. A high score means it thinks hard problems through, from careful analysis to plans that span many steps.</dd></div><div><dt>Visual Intelligence</dt><dd>The ability to understand visual intent and turn ideas, descriptions and plans into coherent, effective visuals. A high score means design ideas come out the way you meant them.</dd></div></dl></details>
<div class="filters" role="group" aria-label="Models">
<div class="filter-types">
<label class="choice"><input type="radio" name="type" value="all" checked><span class="choice-name">All models</span></label>
<label class="choice"><input type="radio" name="type" value="open"><span class="choice-name">Open weights</span><small>The weights can be downloaded.</small></label>
<label class="choice"><input type="radio" name="type" value="local"><span class="choice-name">Local</span><small>Open weights that run on a consumer PC.</small></label>
</div>
<div class="filter-search"><label for="model-search" class="sr-only">Search models</label><input id="model-search" type="search" placeholder="Search models" autocomplete="off" spellcheck="false"><button type="button" class="clear" data-clear hidden>Clear filters</button></div>
</div>
<p class="no-match" data-no-match hidden role="status">No models match these filters. <button type="button" class="clear" data-clear>Clear filters</button></p>
<table class="lb" data-lb>
<thead><tr><th scope="col" class="lb-model">Model</th><th scope="col" class="num" aria-sort="descending"><button type="button" class="sort" data-sort="score" aria-label="Sort by Benchmark score">Score</button></th><th scope="col" class="num"><button type="button" class="sort" data-sort="general" aria-label="Sort by General Intelligence">GI</button></th><th scope="col" class="num"><button type="button" class="sort" data-sort="visual" aria-label="Sort by Visual Intelligence">VI</button></th></tr></thead>
<tbody data-group="complete"><tr data-model="openai/gpt-6-astra" data-open="0" data-local="0" data-score="74.7" data-general="73.4" data-visual="77" data-name="GPT-6 Astra" data-search="GPT-6 Astra"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/gpt-6-astra?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="openai/gpt-6-astra"><span class="mount"><img src="/benchmarks/brands/openai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="GPT-6 Astra">GPT-6 Astra</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/gpt-6-astra?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect GPT-6 Astra on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">74.7</td><td class="num num--gi" data-label="GI">73.4</td><td class="num num--vi" data-label="VI">77.0</td></tr><tr data-model="anthropic/claude-opus-5.5" data-open="0" data-local="0" data-score="72.6" data-general="71.3" data-visual="75" data-name="Claude Opus 5.5" data-search="Claude Opus 5.5"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/claude-opus-5.5?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="anthropic/claude-opus-5.5"><span class="mount"><img src="/benchmarks/brands/anthropic.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Claude Opus 5.5">Claude Opus 5.5</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/claude-opus-5.5?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Claude Opus 5.5 on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">72.6</td><td class="num num--gi" data-label="GI">71.3</td><td class="num num--vi" data-label="VI">75.0</td></tr><tr data-model="anthropic/claude-sonnet-5.5" data-open="0" data-local="0" data-score="70.2" data-general="67" data-visual="76.1" data-name="Claude Sonnet 5.5" data-search="Claude Sonnet 5.5"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/claude-sonnet-5.5?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="anthropic/claude-sonnet-5.5"><span class="mount"><img src="/benchmarks/brands/anthropic.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Claude Sonnet 5.5">Claude Sonnet 5.5</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/claude-sonnet-5.5?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Claude Sonnet 5.5 on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">70.2</td><td class="num num--gi" data-label="GI">67.0</td><td class="num num--vi" data-label="VI">76.1</td></tr><tr data-model="openai/gpt-6.1-sol" data-open="0" data-local="0" data-score="67.8" data-general="64.4" data-visual="74.1" data-name="GPT-6.1 Sol" data-search="GPT-6.1 Sol"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/gpt-6.1-sol?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="openai/gpt-6.1-sol"><span class="mount"><img src="/benchmarks/brands/openai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="GPT-6.1 Sol">GPT-6.1 Sol</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/gpt-6.1-sol?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect GPT-6.1 Sol on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">67.8</td><td class="num num--gi" data-label="GI">64.4</td><td class="num num--vi" data-label="VI">74.1</td></tr><tr data-model="google/gemini-4-argon" data-open="0" data-local="0" data-score="67.4" data-general="63.6" data-visual="74.4" data-name="Gemini 4 Argon" data-search="Gemini 4 Argon"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/gemini-4-argon?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="google/gemini-4-argon"><span class="mount"><img src="/benchmarks/brands/google.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Gemini 4 Argon">Gemini 4 Argon</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/gemini-4-argon?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Gemini 4 Argon on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">67.4</td><td class="num num--gi" data-label="GI">63.6</td><td class="num num--vi" data-label="VI">74.4</td></tr><tr data-model="qwen/qwen3.8-max-0902" data-open="0" data-local="0" data-score="63.3" data-general="62.7" data-visual="64.5" data-name="Qwen3.8-Max-0902" data-search="Qwen3.8-Max-0902"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/qwen3.8-max-0902?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="qwen/qwen3.8-max-0902"><span class="mount"><img src="/benchmarks/brands/qwen.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Qwen3.8-Max-0902">Qwen3.8-Max-0902</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/qwen3.8-max-0902?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Qwen3.8-Max-0902 on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">63.3</td><td class="num num--gi" data-label="GI">62.7</td><td class="num num--vi" data-label="VI">64.5</td></tr><tr data-model="anthropic/claude-fable-5.1" data-open="0" data-local="0" data-score="63" data-general="60.6" data-visual="67.1" data-name="Claude Fable 5.1" data-search="Claude Fable 5.1"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/claude-fable-5.1?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="anthropic/claude-fable-5.1"><span class="mount"><img src="/benchmarks/brands/anthropic.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Claude Fable 5.1">Claude Fable 5.1</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/claude-fable-5.1?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Claude Fable 5.1 on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">63.0</td><td class="num num--gi" data-label="GI">60.6</td><td class="num num--vi" data-label="VI">67.1</td></tr><tr data-model="union-alpha" data-open="0" data-local="0" data-score="62.6" data-general="58.7" data-visual="69.6" data-name="Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)" data-search="Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM) Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM) Pareto Pareto 26.9"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/union-alpha?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="union-alpha"><span class="mount"><img src="/_shared/favicons/union-alpha.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)">Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM)</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a><span class="tag">Router</span></div><div class="lb-links"><a class="sec" href="https://huggingface.co/stealth-model/union-alpha" target="_blank" rel="noopener noreferrer">HF card<span class="sr-only"> · Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM), Our findings on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/union-alpha?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Pareto 26.9 (Union of GPT-6 Astra + DeepSeek + GLM) on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">62.6</td><td class="num num--gi" data-label="GI">58.7</td><td class="num num--vi" data-label="VI">69.6</td></tr><tr data-model="openai/gpt-6.1-terra" data-open="0" data-local="0" data-score="60.4" data-general="58" data-visual="64.9" data-name="GPT-6.1 Terra" data-search="GPT-6.1 Terra"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/gpt-6.1-terra?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="openai/gpt-6.1-terra"><span class="mount"><img src="/benchmarks/brands/openai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="GPT-6.1 Terra">GPT-6.1 Terra</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/gpt-6.1-terra?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect GPT-6.1 Terra on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">60.4</td><td class="num num--gi" data-label="GI">58.0</td><td class="num num--vi" data-label="VI">64.9</td></tr><tr data-model="deepseek/deepseek-v4.1-flash-1017" data-open="1" data-local="0" data-score="59.2" data-general="56.8" data-visual="63.5" data-name="DeepSeek-V4.1-Flash-1017" data-search="DeepSeek-V4.1-Flash-1017"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/deepseek-v4.1-flash-1017?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="deepseek/deepseek-v4.1-flash-1017"><span class="mount"><img src="/benchmarks/brands/deepseek.svg" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="DeepSeek-V4.1-Flash-1017">DeepSeek-V4.1-Flash-1017</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · DeepSeek-V4.1-Flash-1017, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/deepseek-v4.1-flash-1017?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect DeepSeek-V4.1-Flash-1017 on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">59.2</td><td class="num num--gi" data-label="GI">56.8</td><td class="num num--vi" data-label="VI">63.5</td></tr><tr data-model="meta/muse-wonder" data-open="0" data-local="0" data-score="57" data-general="53.1" data-visual="64" data-name="Muse Wonder" data-search="Muse Wonder"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/muse-wonder?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="meta/muse-wonder"><span class="mount"><img src="/benchmarks/brands/meta.svg" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Muse Wonder">Muse Wonder</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/muse-wonder?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Muse Wonder on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">57.0</td><td class="num num--gi" data-label="GI">53.1</td><td class="num num--vi" data-label="VI">64.0</td></tr><tr data-model="moonshotai/kimi-k3.1-pro" data-open="1" data-local="0" data-score="56.7" data-general="53.9" data-visual="62" data-name="Kimi K3.1 Pro" data-search="Kimi K3.1 Pro"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/kimi-k3.1-pro?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="moonshotai/kimi-k3.1-pro"><span class="mount"><img src="/benchmarks/brands/moonshot.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Kimi K3.1 Pro">Kimi K3.1 Pro</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/kimi-k3.1-pro?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Kimi K3.1 Pro on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">56.7</td><td class="num num--gi" data-label="GI">53.9</td><td class="num num--vi" data-label="VI">62.0</td></tr><tr data-model="openai/gpt-5.6-sol" data-open="0" data-local="0" data-score="54.1" data-general="51.9" data-visual="58.2" data-name="GPT-5.6 Sol" data-search="GPT-5.6 Sol"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/gpt-5.6-sol?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="openai/gpt-5.6-sol"><span class="mount"><img src="/benchmarks/brands/openai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="GPT-5.6 Sol">GPT-5.6 Sol</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/gpt-5.6-sol?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect GPT-5.6 Sol on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">54.1</td><td class="num num--gi" data-label="GI">51.9</td><td class="num num--vi" data-label="VI">58.2</td></tr><tr data-model="openai/gpt-6-sol" data-open="0" data-local="0" data-score="51.7" data-general="46.1" data-visual="61.6" data-name="GPT-6 Sol" data-search="GPT-6 Sol"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/gpt-6-sol?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="openai/gpt-6-sol"><span class="mount"><img src="/benchmarks/brands/openai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="GPT-6 Sol">GPT-6 Sol</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/gpt-6-sol?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect GPT-6 Sol on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">51.7</td><td class="num num--gi" data-label="GI">46.1</td><td class="num num--vi" data-label="VI">61.6</td></tr><tr data-model="qwen/qwen3.8-27B" data-open="1" data-local="1" data-score="51.7" data-general="46.2" data-visual="61.4" data-name="Qwen3.8-27B" data-search="Qwen3.8-27B"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/qwen3.8-27b?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="qwen/qwen3.8-27B"><span class="mount"><img src="/benchmarks/brands/qwen.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Qwen3.8-27B">Qwen3.8-27B</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/Qwen/Qwen3.8-27B" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · Qwen3.8-27B, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/qwen3.8-27b?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Qwen3.8-27B on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">51.7</td><td class="num num--gi" data-label="GI">46.2</td><td class="num num--vi" data-label="VI">61.4</td></tr><tr data-model="moonshotai/kimi-k3" data-open="1" data-local="0" data-score="50.6" data-general="44.4" data-visual="61.4" data-name="Kimi K3" data-search="Kimi K3"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/kimi-k3?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="moonshotai/kimi-k3"><span class="mount"><img src="/benchmarks/brands/moonshot.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Kimi K3">Kimi K3</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/moonshotai/Kimi-K3" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · Kimi K3, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/kimi-k3?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Kimi K3 on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">50.6</td><td class="num num--gi" data-label="GI">44.4</td><td class="num num--vi" data-label="VI">61.4</td></tr><tr data-model="deepseek/deepseek-v4.1-flash" data-open="1" data-local="0" data-score="50" data-general="48.5" data-visual="53" data-name="DeepSeek-V4.1-Flash" data-search="DeepSeek-V4.1-Flash"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/deepseek-v4.1-flash?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="deepseek/deepseek-v4.1-flash"><span class="mount"><img src="/benchmarks/brands/deepseek.svg" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="DeepSeek-V4.1-Flash">DeepSeek-V4.1-Flash</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · DeepSeek-V4.1-Flash, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/deepseek-v4.1-flash?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect DeepSeek-V4.1-Flash on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">50.0</td><td class="num num--gi" data-label="GI">48.5</td><td class="num num--vi" data-label="VI">53.0</td></tr><tr data-model="anthropic/claude-haiku-5.5" data-open="0" data-local="0" data-score="49.5" data-general="47.6" data-visual="52.8" data-name="Claude Haiku 5.5" data-search="Claude Haiku 5.5"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/claude-haiku-5.5?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="anthropic/claude-haiku-5.5"><span class="mount"><img src="/benchmarks/brands/anthropic.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Claude Haiku 5.5">Claude Haiku 5.5</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/claude-haiku-5.5?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Claude Haiku 5.5 on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">49.5</td><td class="num num--gi" data-label="GI">47.6</td><td class="num num--vi" data-label="VI">52.8</td></tr><tr data-model="ox-alpha" data-open="1" data-local="1" data-score="48.5" data-general="42.3" data-visual="59.4" data-name="GLM-5.3-Flash" data-search="GLM-5.3-Flash GLM-5.3-Flash ox-alpha stealth/ox-alpha GLM-5.3-Flash GLM 5.3 Flash glm-5.3-flash glm-5-3-flash z-ai/glm-5.3-flash Z.ai GLM-5.3-Flash"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/glm-5.3-flash?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="ox-alpha"><span class="mount"><img src="/_shared/favicons/ox-alpha.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="GLM-5.3-Flash">GLM-5.3-Flash</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/zai-org/GLM-5.3-Flash" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · GLM-5.3-Flash, Official weights on Hugging Face</span></a><a class="sec" href="https://huggingface.co/stealth-model/GLM-5.3-Flash-GGUF" target="_blank" rel="noopener noreferrer">GGUF<span class="sr-only"> · GGUF weights on Hugging Face</span></a><a class="sec" href="https://huggingface.co/stealth-model/ox-alpha" target="_blank" rel="noopener noreferrer">HF card<span class="sr-only"> · GLM-5.3-Flash, Our findings on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/glm-5.3-flash?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect GLM-5.3-Flash on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">48.5</td><td class="num num--gi" data-label="GI">42.3</td><td class="num num--vi" data-label="VI">59.4</td></tr><tr data-model="google/gemini-3.8-flash" data-open="0" data-local="0" data-score="47.9" data-general="43.9" data-visual="55.2" data-name="Gemini 3.8 Flash" data-search="Gemini 3.8 Flash"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/gemini-3.8-flash?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="google/gemini-3.8-flash"><span class="mount"><img src="/benchmarks/brands/google.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Gemini 3.8 Flash">Gemini 3.8 Flash</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/gemini-3.8-flash?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Gemini 3.8 Flash on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">47.9</td><td class="num num--gi" data-label="GI">43.9</td><td class="num num--vi" data-label="VI">55.2</td></tr><tr data-model="xiaomi/mimo-v2.6-pro" data-open="1" data-local="0" data-score="47.9" data-general="44.3" data-visual="54.3" data-name="MiMo-V2.6-Pro" data-search="MiMo-V2.6-Pro"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/mimo-v2.6-pro?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="xiaomi/mimo-v2.6-pro"><span class="mount"><img src="/benchmarks/brands/xiaomi.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="MiMo-V2.6-Pro">MiMo-V2.6-Pro</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Pro-RL" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · MiMo-V2.6-Pro, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/mimo-v2.6-pro?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect MiMo-V2.6-Pro on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">47.9</td><td class="num num--gi" data-label="GI">44.3</td><td class="num num--vi" data-label="VI">54.3</td></tr><tr data-model="x-ai/grok-4.7" data-open="0" data-local="0" data-score="47.2" data-general="48.4" data-visual="44.8" data-name="Grok 4.7" data-search="Grok 4.7"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/grok-4.7?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="x-ai/grok-4.7"><span class="mount"><img src="/benchmarks/brands/x-ai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Grok 4.7">Grok 4.7</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/grok-4.7?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Grok 4.7 on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">47.2</td><td class="num num--gi" data-label="GI">48.4</td><td class="num num--vi" data-label="VI">44.8</td></tr><tr data-model="qwen/qwen3.8-flash" data-open="1" data-local="1" data-score="46.6" data-general="38.5" data-visual="60.3" data-name="Qwen3.8-Flash" data-search="Qwen3.8-Flash"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/qwen3.8-flash?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="qwen/qwen3.8-flash"><span class="mount"><img src="/benchmarks/brands/qwen.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Qwen3.8-Flash">Qwen3.8-Flash</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/Qwen/Qwen3.8-Flash-Next" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · Qwen3.8-Flash, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/qwen3.8-flash?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Qwen3.8-Flash on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">46.6</td><td class="num num--gi" data-label="GI">38.5</td><td class="num num--vi" data-label="VI">60.3</td></tr><tr data-model="minimax/minimax-m3.1-flash" data-open="1" data-local="0" data-score="45.9" data-general="35.8" data-visual="62.6" data-name="MiniMax M3.1 Flash" data-search="MiniMax M3.1 Flash"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/minimax-m3.1-flash?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="minimax/minimax-m3.1-flash"><span class="mount"><img src="/benchmarks/brands/minimax.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="MiniMax M3.1 Flash">MiniMax M3.1 Flash</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/minimax-m3.1-flash?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect MiniMax M3.1 Flash on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">45.9</td><td class="num num--gi" data-label="GI">35.8</td><td class="num num--vi" data-label="VI">62.6</td></tr><tr data-model="openai/gpt-5.6-terra" data-open="0" data-local="0" data-score="45.8" data-general="41.9" data-visual="52.7" data-name="GPT-5.6 Terra" data-search="GPT-5.6 Terra"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/gpt-5.6-terra?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="openai/gpt-5.6-terra"><span class="mount"><img src="/benchmarks/brands/openai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="GPT-5.6 Terra">GPT-5.6 Terra</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/gpt-5.6-terra?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect GPT-5.6 Terra on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">45.8</td><td class="num num--gi" data-label="GI">41.9</td><td class="num num--vi" data-label="VI">52.7</td></tr><tr data-model="space-bunny-alpha" data-open="0" data-local="0" data-score="45.5" data-general="35.7" data-visual="64.4" data-name="MiniMax M3.1 Flash (early version)" data-search="MiniMax M3.1 Flash (early version) MiniMax M3.1 Flash (early version) space-bunny-alpha Space Bunny Free Space Bunny stealth/space-bunny-alpha"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/space-bunny-alpha?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="space-bunny-alpha"><span class="mount"><img src="/_shared/favicons/space-bunny-alpha.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="MiniMax M3.1 Flash (early version)">MiniMax M3.1 Flash (early version)</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/stealth-model/space-bunny-alpha" target="_blank" rel="noopener noreferrer">HF card<span class="sr-only"> · MiniMax M3.1 Flash (early version), Our findings on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/space-bunny-alpha?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect MiniMax M3.1 Flash (early version) on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">45.5</td><td class="num num--gi" data-label="GI">35.7</td><td class="num num--vi" data-label="VI">64.4</td></tr><tr data-model="z-ai/glm-5.3" data-open="1" data-local="0" data-score="45.1" data-general="37.6" data-visual="57.2" data-name="GLM-5.3" data-search="GLM-5.3"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/glm-5.3?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="z-ai/glm-5.3"><span class="mount"><img src="/benchmarks/brands/z-ai.svg" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="GLM-5.3">GLM-5.3</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/zai-org/GLM-5.3" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · GLM-5.3, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/glm-5.3?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect GLM-5.3 on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">45.1</td><td class="num num--gi" data-label="GI">37.6</td><td class="num num--vi" data-label="VI">57.2</td></tr><tr data-model="xiaomi/mimo-v2.6-flash" data-open="1" data-local="0" data-score="44.6" data-general="41" data-visual="49.8" data-name="MiMo-V2.6-Flash" data-search="MiMo-V2.6-Flash"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/mimo-v2.6-flash?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="xiaomi/mimo-v2.6-flash"><span class="mount"><img src="/benchmarks/brands/xiaomi.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="MiMo-V2.6-Flash">MiMo-V2.6-Flash</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-RL" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · MiMo-V2.6-Flash, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/mimo-v2.6-flash?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect MiMo-V2.6-Flash on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">44.6</td><td class="num num--gi" data-label="GI">41.0</td><td class="num num--vi" data-label="VI">49.8</td></tr><tr data-model="meta/muse-spark-open" data-open="1" data-local="0" data-score="44.1" data-general="42.6" data-visual="46.6" data-name="Muse Spark Open" data-search="Muse Spark Open"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/muse-spark-open?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="meta/muse-spark-open"><span class="mount"><img src="/benchmarks/brands/meta.svg" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Muse Spark Open">Muse Spark Open</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/muse-spark-open?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Muse Spark Open on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">44.1</td><td class="num num--gi" data-label="GI">42.6</td><td class="num num--vi" data-label="VI">46.6</td></tr><tr data-model="tencent/hy4-preview" data-open="1" data-local="0" data-score="43.9" data-general="41.3" data-visual="48.6" data-name="Hy4 preview" data-search="Hy4 preview"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/hy4-preview?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="tencent/hy4-preview"><span class="mount"><img src="/benchmarks/brands/tencent.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Hy4 preview">Hy4 preview</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/tencent/Hy4-preview" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · Hy4 preview, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/hy4-preview?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Hy4 preview on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">43.9</td><td class="num num--gi" data-label="GI">41.3</td><td class="num num--vi" data-label="VI">48.6</td></tr><tr data-model="mistralai/mistral-large-4-0" data-open="0" data-local="0" data-score="43.3" data-general="34.4" data-visual="58" data-name="Mistral Large 4" data-search="Mistral Large 4"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/mistral-large-4-0?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="mistralai/mistral-large-4-0"><span class="mount"><img src="/benchmarks/brands/mistral.svg" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Mistral Large 4">Mistral Large 4</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/mistral-large-4-0?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Mistral Large 4 on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">43.3</td><td class="num num--gi" data-label="GI">34.4</td><td class="num num--vi" data-label="VI">58.0</td></tr><tr data-model="inclusionai/ling-3.1-flash" data-open="1" data-local="0" data-score="42.2" data-general="35.5" data-visual="51.2" data-name="Ling 3.1 Flash" data-search="Ling 3.1 Flash"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/ling-3.1-flash?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="inclusionai/ling-3.1-flash"><span class="mount"><img src="/benchmarks/brands/inclusionai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Ling 3.1 Flash">Ling 3.1 Flash</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/inclusionAI/Ling-3.1-flash" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · Ling 3.1 Flash, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/ling-3.1-flash?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Ling 3.1 Flash on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">42.2</td><td class="num num--gi" data-label="GI">35.5</td><td class="num num--vi" data-label="VI">51.2</td></tr><tr data-model="deepseek/deepseek-v4-pro-0813" data-open="1" data-local="0" data-score="41.8" data-general="41.6" data-visual="42.3" data-name="DeepSeek-V4-Pro-0813" data-search="DeepSeek-V4-Pro-0813"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/deepseek-v4-pro-0813?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="deepseek/deepseek-v4-pro-0813"><span class="mount"><img src="/benchmarks/brands/deepseek.svg" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="DeepSeek-V4-Pro-0813">DeepSeek-V4-Pro-0813</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro-0813" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · DeepSeek-V4-Pro-0813, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/deepseek-v4-pro-0813?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect DeepSeek-V4-Pro-0813 on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">41.8</td><td class="num num--gi" data-label="GI">41.6</td><td class="num num--vi" data-label="VI">42.3</td></tr><tr data-model="openai/gpt-6.1-terra-low" data-open="0" data-local="0" data-score="41.3" data-general="34.3" data-visual="53.1" data-name="GPT-6.1 Terra Low" data-search="GPT-6.1 Terra Low"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/gpt-6.1-terra-low?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="openai/gpt-6.1-terra-low"><span class="mount"><img src="/benchmarks/brands/openai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="GPT-6.1 Terra Low">GPT-6.1 Terra Low</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/gpt-6.1-terra-low?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect GPT-6.1 Terra Low on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">41.3</td><td class="num num--gi" data-label="GI">34.3</td><td class="num num--vi" data-label="VI">53.1</td></tr><tr data-model="meta/muse-spark-1.3" data-open="0" data-local="0" data-score="39.8" data-general="38.3" data-visual="42.3" data-name="Muse Spark 1.3" data-search="Muse Spark 1.3"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/muse-spark-1.3?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="meta/muse-spark-1.3"><span class="mount"><img src="/benchmarks/brands/meta.svg" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Muse Spark 1.3">Muse Spark 1.3</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/muse-spark-1.3?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Muse Spark 1.3 on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">39.8</td><td class="num num--gi" data-label="GI">38.3</td><td class="num num--vi" data-label="VI">42.3</td></tr><tr data-model="openai/gpt-6-luna" data-open="0" data-local="0" data-score="39.1" data-general="31.3" data-visual="52" data-name="GPT-6 Luna" data-search="GPT-6 Luna"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/gpt-6-luna?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="openai/gpt-6-luna"><span class="mount"><img src="/benchmarks/brands/openai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="GPT-6 Luna">GPT-6 Luna</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/gpt-6-luna?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect GPT-6 Luna on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">39.1</td><td class="num num--gi" data-label="GI">31.3</td><td class="num num--vi" data-label="VI">52.0</td></tr><tr data-model="stepfun/step-5-preview" data-open="1" data-local="0" data-score="39" data-general="37.1" data-visual="41.9" data-name="Step 5 Preview" data-search="Step 5 Preview"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/step-5-preview?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="stepfun/step-5-preview"><span class="mount"><i class="dot" aria-hidden="true"></i></span><span class="lname" title="Step 5 Preview">Step 5 Preview</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/stepfun-ai/Step-5-Preview" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · Step 5 Preview, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/step-5-preview?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Step 5 Preview on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">39.0</td><td class="num num--gi" data-label="GI">37.1</td><td class="num num--vi" data-label="VI">41.9</td></tr><tr data-model="openai/gpt-5.6-luna" data-open="0" data-local="0" data-score="37.8" data-general="30.4" data-visual="50.2" data-name="GPT-5.6 Luna" data-search="GPT-5.6 Luna"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/gpt-5.6-luna?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="openai/gpt-5.6-luna"><span class="mount"><img src="/benchmarks/brands/openai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="GPT-5.6 Luna">GPT-5.6 Luna</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/gpt-5.6-luna?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect GPT-5.6 Luna on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">37.8</td><td class="num num--gi" data-label="GI">30.4</td><td class="num num--vi" data-label="VI">50.2</td></tr><tr data-model="openai/gpt-5.5" data-open="0" data-local="0" data-score="36.5" data-general="36.6" data-visual="36.3" data-name="GPT-5.5" data-search="GPT-5.5"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/gpt-5.5?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="openai/gpt-5.5"><span class="mount"><img src="/benchmarks/brands/openai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="GPT-5.5">GPT-5.5</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/gpt-5.5?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect GPT-5.5 on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">36.5</td><td class="num num--gi" data-label="GI">36.6</td><td class="num num--vi" data-label="VI">36.3</td></tr><tr data-model="minimax/minimax-m3" data-open="1" data-local="0" data-score="34.6" data-general="28" data-visual="45.3" data-name="MiniMax M3" data-search="MiniMax M3"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/minimax-m3?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="minimax/minimax-m3"><span class="mount"><img src="/benchmarks/brands/minimax.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="MiniMax M3">MiniMax M3</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/MiniMaxAI/MiniMax-M3" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · MiniMax M3, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/minimax-m3?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect MiniMax M3 on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">34.6</td><td class="num num--gi" data-label="GI">28.0</td><td class="num num--vi" data-label="VI">45.3</td></tr><tr data-model="qwen/qwen3.6-27b" data-open="1" data-local="1" data-score="32.4" data-general="25.6" data-visual="43.9" data-name="Qwen3.6-27B" data-search="Qwen3.6-27B"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/qwen3.6-27b?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="qwen/qwen3.6-27b"><span class="mount"><img src="/benchmarks/brands/qwen.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Qwen3.6-27B">Qwen3.6-27B</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/Qwen/Qwen3.6-27B" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · Qwen3.6-27B, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/qwen3.6-27b?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Qwen3.6-27B on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">32.4</td><td class="num num--gi" data-label="GI">25.6</td><td class="num num--vi" data-label="VI">43.9</td></tr><tr data-model="reflection/beam-1.0-501b-a23b" data-open="1" data-local="0" data-score="31.9" data-general="30.9" data-visual="33.8" data-name="Beam 1.0 501B A23B" data-search="Beam 1.0 501B A23B"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/beam-1.0-501b-a23b?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="reflection/beam-1.0-501b-a23b"><span class="mount"><img src="/benchmarks/brands/reflection-ai.svg" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Beam 1.0 501B A23B">Beam 1.0 501B A23B</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/reflection/Beam-1.0-501B-A23B" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · Beam 1.0 501B A23B, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/beam-1.0-501b-a23b?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Beam 1.0 501B A23B on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">31.9</td><td class="num num--gi" data-label="GI">30.9</td><td class="num num--vi" data-label="VI">33.8</td></tr><tr data-model="thinkingmachines/inkling-small" data-open="1" data-local="0" data-score="27.6" data-general="25.5" data-visual="31.1" data-name="Inkling-Small" data-search="Inkling-Small"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/inkling-small?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="thinkingmachines/inkling-small"><span class="mount"><i class="dot" aria-hidden="true"></i></span><span class="lname" title="Inkling-Small">Inkling-Small</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/thinkingmachines/Inkling-Small" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · Inkling-Small, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/inkling-small?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Inkling-Small on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">27.6</td><td class="num num--gi" data-label="GI">25.5</td><td class="num num--vi" data-label="VI">31.1</td></tr><tr data-model="google/gemma-4-31b-it" data-open="1" data-local="1" data-score="25.8" data-general="19.1" data-visual="36.7" data-name="Gemma 4 31B" data-search="Gemma 4 31B"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/gemma-4-31b-it?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="google/gemma-4-31b-it"><span class="mount"><img src="/benchmarks/brands/google.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Gemma 4 31B">Gemma 4 31B</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/google/gemma-4-31B-it" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · Gemma 4 31B, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/gemma-4-31b-it?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Gemma 4 31B on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">25.8</td><td class="num num--gi" data-label="GI">19.1</td><td class="num num--vi" data-label="VI">36.7</td></tr><tr data-model="qwen/qwen3.6-35b-a3b" data-open="1" data-local="1" data-score="20.8" data-general="12.7" data-visual="34" data-name="Qwen3.6-35B-A3B" data-search="Qwen3.6-35B-A3B"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/qwen3.6-35b-a3b?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="qwen/qwen3.6-35b-a3b"><span class="mount"><img src="/benchmarks/brands/qwen.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Qwen3.6-35B-A3B">Qwen3.6-35B-A3B</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/Qwen/Qwen3.6-35B-A3B" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · Qwen3.6-35B-A3B, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/qwen3.6-35b-a3b?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Qwen3.6-35B-A3B on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">20.8</td><td class="num num--gi" data-label="GI">12.7</td><td class="num num--vi" data-label="VI">34.0</td></tr><tr data-model="iceman-contributor" data-open="0" data-local="0" data-score="19.2" data-general="11.7" data-visual="31" data-name="Iceman Contributor" data-search="Iceman Contributor Iceman Iceman Contributor Iceman Contributor Free iceman-contributor-free opencode-go/iceman-contributor-free opencode/iceman-contributor-free"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/iceman-contributor?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="iceman-contributor"><span class="mount"><img src="/_shared/favicons/iceman-contributor.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Iceman Contributor">Iceman Contributor</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/stealth-model/iceman-contributor" target="_blank" rel="noopener noreferrer">HF card<span class="sr-only"> · Iceman Contributor, Our findings on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/iceman-contributor?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Iceman Contributor on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">19.2</td><td class="num num--gi" data-label="GI">11.7</td><td class="num num--vi" data-label="VI">31.0</td></tr><tr data-model="glyph-cluster" data-open="0" data-local="0" data-score="18.9" data-general="12" data-visual="29.8" data-name="Glyph Cluster" data-search="Glyph Cluster stealth/glyph-cluster"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/glyph-cluster?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="glyph-cluster"><span class="mount"><img src="/_shared/favicons/glyph-cluster.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Glyph Cluster">Glyph Cluster</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/stealth-model/glyph-cluster" target="_blank" rel="noopener noreferrer">HF card<span class="sr-only"> · Glyph Cluster, Our findings on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/glyph-cluster?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Glyph Cluster on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">18.9</td><td class="num num--gi" data-label="GI">12.0</td><td class="num num--vi" data-label="VI">29.8</td></tr><tr data-model="nvidia/nemotron-3-ultra-550b-a55b" data-open="1" data-local="0" data-score="16.7" data-general="12.2" data-visual="26.5" data-name="Nemotron 3 Ultra" data-search="Nemotron 3 Ultra"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/nemotron-3-ultra-550b-a55b?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="nvidia/nemotron-3-ultra-550b-a55b"><span class="mount"><img src="/benchmarks/brands/nvidia.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Nemotron 3 Ultra">Nemotron 3 Ultra</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · Nemotron 3 Ultra, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/nemotron-3-ultra-550b-a55b?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Nemotron 3 Ultra on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">16.7</td><td class="num num--gi" data-label="GI">12.2</td><td class="num num--vi" data-label="VI">26.5</td></tr><tr data-model="google/gemma-4-26b-a4b-it" data-open="1" data-local="1" data-score="16.5" data-general="9.4" data-visual="28.9" data-name="Gemma 4 26B A4B" data-search="Gemma 4 26B A4B"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/gemma-4-26b-a4b-it?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="google/gemma-4-26b-a4b-it"><span class="mount"><img src="/benchmarks/brands/google.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Gemma 4 26B A4B">Gemma 4 26B A4B</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/google/gemma-4-26B-A4B-it" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · Gemma 4 26B A4B, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/gemma-4-26b-a4b-it?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Gemma 4 26B A4B on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">16.5</td><td class="num num--gi" data-label="GI">9.4</td><td class="num num--vi" data-label="VI">28.9</td></tr><tr data-model="microsoft/mai-thinking-1" data-open="0" data-local="0" data-score="14.1" data-general="7.4" data-visual="24.6" data-name="MAI-Thinking-1" data-search="MAI-Thinking-1"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/mai-thinking-1?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="microsoft/mai-thinking-1"><span class="mount"><img src="/benchmarks/brands/microsoft-ai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="MAI-Thinking-1">MAI-Thinking-1</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/mai-thinking-1?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect MAI-Thinking-1 on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">14.1</td><td class="num num--gi" data-label="GI">7.4</td><td class="num num--vi" data-label="VI">24.6</td></tr><tr data-model="qwen/qwen3.5-9b" data-open="1" data-local="1" data-score="10.9" data-general="7.9" data-visual="15.6" data-name="Qwen3.5-9B" data-search="Qwen3.5-9B"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/qwen3.5-9b?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="qwen/qwen3.5-9b"><span class="mount"><img src="/benchmarks/brands/qwen.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Qwen3.5-9B">Qwen3.5-9B</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="lb-links"><a class="sec" href="https://huggingface.co/Qwen/Qwen3.5-9B" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · Qwen3.5-9B, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/qwen3.5-9b?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Qwen3.5-9B on stealthmodels.com">Inspect</a></div></th><td class="num num--score" data-label="Score">10.9</td><td class="num num--gi" data-label="GI">7.9</td><td class="num num--vi" data-label="VI">15.6</td></tr></tbody>
<tbody data-group="partial"><tr class="lb-sep"><th colspan="4" scope="colgroup">Partial results</th></tr><tr data-model="pixel-canary" data-open="0" data-local="0" data-score="49.5" data-general="40.4" data-visual="59.6" data-name="Callstack Apex" data-search="Callstack Apex Callstack Apex pixel-canary stealth/pixel-canary Apex Callstack Apex callstack/apex apex-20261001"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/pixel-canary?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="pixel-canary"><span class="mount"><img src="/_shared/favicons/pixel-canary.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Callstack Apex">Callstack Apex</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a><span class="tag">Partial coverage</span></div><div class="lb-links"><a class="sec" href="https://huggingface.co/stealth-model/pixel-canary" target="_blank" rel="noopener noreferrer">HF card<span class="sr-only"> · Callstack Apex, Our findings on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/pixel-canary?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Callstack Apex on stealthmodels.com">Inspect</a></div><details class="lb-cover"><summary>Partial coverage</summary><p>Not scored in Foresight</p></details></th><td class="num num--score" data-label="Score">49.5</td><td class="num num--gi" data-label="GI">40.4</td><td class="num num--vi" data-label="VI">59.6</td></tr><tr data-model="fledge-alpha" data-open="0" data-local="0" data-score="35.8" data-general="27.1" data-visual="45.2" data-name="Fledge Alpha" data-search="Fledge Alpha fledge-alpha fledge-alpha-free fledge alpha opencode/fledge-alpha-free"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/fledge-alpha?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="fledge-alpha"><span class="mount"><img src="/_shared/favicons/fledge-alpha.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Fledge Alpha">Fledge Alpha</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a><span class="tag">Suspected router</span><span class="tag">Partial coverage</span></div><div class="lb-links"><a class="sec" href="https://huggingface.co/stealth-model/fledge-alpha" target="_blank" rel="noopener noreferrer">HF card<span class="sr-only"> · Fledge Alpha, Our findings on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/fledge-alpha?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Fledge Alpha on stealthmodels.com">Inspect</a></div><details class="lb-cover"><summary>Partial coverage</summary><p>Not scored in Foresight, Latch</p></details></th><td class="num num--score" data-label="Score">35.8</td><td class="num num--gi" data-label="GI">27.1</td><td class="num num--vi" data-label="VI">45.2</td></tr><tr data-model="owl-alpha" data-open="1" data-local="0" data-score="25.5" data-general="17.2" data-visual="34.4" data-name="LongCat-2.0" data-search="LongCat-2.0 LongCat-2.0 openrouter/owl-alpha LongCat-2.0 LongCat-2.0-Preview meituan/longcat-2.0"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/owl-alpha?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="owl-alpha"><span class="mount"><img src="/_shared/favicons/owl-alpha.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="LongCat-2.0">LongCat-2.0</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a><span class="tag">Partial coverage</span></div><div class="lb-links"><a class="sec" href="https://huggingface.co/meituan-longcat/LongCat-2.0" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · LongCat-2.0, Official weights on Hugging Face</span></a><a class="sec" href="https://huggingface.co/stealth-model/owl-alpha" target="_blank" rel="noopener noreferrer">HF card<span class="sr-only"> · LongCat-2.0, Our findings on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/owl-alpha?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect LongCat-2.0 on stealthmodels.com">Inspect</a></div><details class="lb-cover"><summary>Partial coverage</summary><p>Not scored in Foresight, Latch</p></details></th><td class="num num--score" data-label="Score">25.5</td><td class="num num--gi" data-label="GI">17.2</td><td class="num num--vi" data-label="VI">34.4</td></tr><tr data-model="Aleph-Alpha/Kolibri-1" data-open="1" data-local="1" data-score="13.6" data-general="12.3" data-visual="15" data-name="Kolibri-1" data-search="Kolibri-1"><th scope="row" class="lb-model"><div class="lb-id"><a class="lockup lockup--s" href="https://stealthmodels.com/benchmarks/models/kolibri-1?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="Aleph-Alpha/Kolibri-1"><span class="mount"><img src="/benchmarks/brands/aleph-alpha.svg" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Kolibri-1">Kolibri-1</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a><span class="tag">Partial coverage</span></div><div class="lb-links"><a class="sec" href="https://huggingface.co/Aleph-Alpha/Kolibri-1" target="_blank" rel="noopener noreferrer">Weights<span class="sr-only"> · Kolibri-1, Official weights on Hugging Face</span></a><a class="lb-inspect" href="https://stealthmodels.com/benchmarks/models/kolibri-1?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" aria-label="Inspect Kolibri-1 on stealthmodels.com">Inspect</a></div><details class="lb-cover"><summary>Partial coverage</summary><p>Not scored in Foresight, Latch</p></details></th><td class="num num--score" data-label="Score">13.6</td><td class="num num--gi" data-label="GI">12.3</td><td class="num num--vi" data-label="VI">15.0</td></tr></tbody>
</table>
<p class="lb-more"><button type="button" class="btn btn-outline more" data-more aria-expanded="false" data-all="Show all results" data-fewer="Show fewer results" hidden>Show all results</button><a class="quiet" href="https://stealthmodels.com/benchmarks?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">View the full results <span aria-hidden="true">→</span></a></p>
</div>
</section>
<section class="evidence" id="evidence" aria-labelledby="evidence-title">
<div class="wrap evidence-head">
<h2 id="evidence-title">See the evidence</h2>
<p class="section-lead">Each benchmark below shows its highest and lowest scoring result, so you can judge the gap yourself.</p>
</div>
<nav class="chapter-index" aria-label="What each benchmark tests" data-ci-nav>
<div class="wrap ci-wrap">
<div class="ci-desktop"><span class="ci-group">General Intelligence</span><a href="#newtons-mirror" data-ci="newtons-mirror">Newton&#39;s Mirror</a><a href="#ground-truth" data-ci="ground-truth">Ground Truth</a><a href="#latch" data-ci="latch">Latch</a><a href="#foresight" data-ci="foresight">Foresight</a><span class="ci-group">Visual Intelligence</span><a href="#mona-lisa" data-ci="mona-lisa">Mona Lisa</a><a href="#selfie" data-ci="selfie">Selfie</a><a href="#beach" data-ci="beach">Beach</a><a href="#raccoon" data-ci="raccoon">Raccoon</a></div>
<div class="ci-mobile"><span class="ci-current" data-ci-current aria-live="polite">Newton&#39;s Mirror</span><button type="button" class="ci-btn" aria-expanded="false" aria-controls="ci-menu">Chapters</button></div>
</div>
<div class="ci-menu" id="ci-menu" hidden><ol><li><a href="#newtons-mirror" data-ci="newtons-mirror">Newton&#39;s Mirror</a></li><li><a href="#ground-truth" data-ci="ground-truth">Ground Truth</a></li><li><a href="#latch" data-ci="latch">Latch</a></li><li><a href="#foresight" data-ci="foresight">Foresight</a></li><li><a href="#mona-lisa" data-ci="mona-lisa">Mona Lisa</a></li><li><a href="#selfie" data-ci="selfie">Selfie</a></li><li><a href="#beach" data-ci="beach">Beach</a></li><li><a href="#raccoon" data-ci="raccoon">Raccoon</a></li></ol></div>
</nav>
<article class="chapter chapter--newtons-mirror" id="newtons-mirror" aria-labelledby="newtons-mirror-title" data-task="newtons-mirror" data-group="newtons-mirror" data-kind="svg"><div class="wrap"><div class="chapter-head"><h2 id="newtons-mirror-title"><span class="chapter-name">Newton&#39;s Mirror</span><span class="sr-only">: </span><span class="question">Can a model trace light through a hall of mirrors in its head, with no code to do the maths?</span></h2><p class="meaning">A high score means the physical state, geometry and repeated reflections remain consistent.</p></div><details class="reference"><summary>Reference</summary><div class="ref-media"><img src="https://stealthmodels.com/benchmarks/media/5b130a89c0822a26.webp" alt="Reference for Newton&#39;s Mirror" width="1600" height="1200" loading="lazy" decoding="async"></div></details><div class="pair-slot"><div class="pair"><figure class="plate" data-slot="high" data-result="openai-gpt-6-astra-svg-newtons-mirror-5a00e74f7bfe6e8e"><p class="plate-label">Highest score</p><div class="plate-id"><a class="lockup lockup--m" href="https://stealthmodels.com/benchmarks/models/gpt-6-astra?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="openai/gpt-6-astra"><span class="mount"><img src="/benchmarks/brands/openai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="GPT-6 Astra">GPT-6 Astra</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="plate-media"><img src="https://stealthmodels.com/benchmarks/media/3373cf328ebe8292.png" alt="GPT-6 Astra: Newton&#39;s Mirror benchmark result" width="1600" height="1416" loading="lazy" decoding="async"></div><figcaption><p class="plate-score"><span class="ps-label">Benchmark score:</span> <span class="ps-num">70.6</span> <span class="ps-of">out of 100</span></p><p class="plate-meta"><span>Recorded 30 September 2026</span><a class="plain" href="https://stealthmodels.com/benchmarks/models/gpt-6-astra/newtons-mirror?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">View result<span class="sr-only"> · GPT-6 Astra, Newton&#39;s Mirror</span></a></p></figcaption></figure><figure class="plate" data-slot="low" data-result="microsoft-mai-thinking-1-svg-newtons-mirror-5a00e74f7bfe6e8e"><p class="plate-label">Lowest score</p><div class="plate-id"><a class="lockup lockup--m" href="https://stealthmodels.com/benchmarks/models/mai-thinking-1?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="microsoft/mai-thinking-1"><span class="mount"><img src="/benchmarks/brands/microsoft-ai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="MAI-Thinking-1">MAI-Thinking-1</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="plate-media"><img src="https://stealthmodels.com/benchmarks/media/88b2bb766d8c60a5.png" alt="MAI-Thinking-1: Newton&#39;s Mirror benchmark result" width="1600" height="900" loading="lazy" decoding="async"></div><figcaption><p class="plate-score"><span class="ps-label">Benchmark score:</span> <span class="ps-num">8.0</span> <span class="ps-of">out of 100</span></p><p class="plate-meta"><span>Recorded 9 October 2026</span><a class="plain" href="https://stealthmodels.com/benchmarks/models/mai-thinking-1/newtons-mirror?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">View result<span class="sr-only"> · MAI-Thinking-1, Newton&#39;s Mirror</span></a></p></figcaption></figure></div></div><div class="film-row film-row--right" data-film-group="newtons-mirror" hidden><div class="film-slot"><div class="film film--tall" data-film="FtB-aezSKN8" data-aspect="9:16" data-public-from="2026-11-05T23:00:00Z" hidden><div class="film-copy"><p class="film-label">Spatial reasoning: one sphere in a hall of mirrors</p><p class="film-yt"><a href="https://www.youtube.com/watch?v=FtB-aezSKN8" target="_blank" rel="noopener noreferrer">Watch on YouTube<span class="sr-only"> · Spatial reasoning: one sphere in a hall of mirrors</span></a></p></div><div class="film-frame"><a class="film-poster" href="https://www.youtube.com/watch?v=FtB-aezSKN8" target="_blank" rel="noopener noreferrer" data-film-play aria-label="Watch: Spatial reasoning: one sphere in a hall of mirrors"><img data-src="/assets/films/FtB-aezSKN8.webp" alt="" width="360" height="640" decoding="async"><span class="film-play" aria-hidden="true"></span></a></div></div></div></div><p class="chapter-links"><a class="quiet" href="https://stealthmodels.com/stealthmark?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space#newtons-mirror" target="_blank" rel="noopener noreferrer">All Newton&#39;s Mirror results <span aria-hidden="true">→</span></a></p></div></article>
<article class="chapter chapter--ground-truth is-portrait" id="ground-truth" aria-labelledby="ground-truth-title" data-task="ground-truth" data-group="ground-truth" data-kind="svg"><div class="wrap"><div class="chapter-head"><h2 id="ground-truth-title"><span class="chapter-name">Ground Truth</span><span class="sr-only">: </span><span class="question">Can a model reason through a scene where every detail depends on another?</span></h2><p class="meaning">A high score means the model follows how placement, materials and light affect one another.</p></div><details class="reference"><summary>Reference</summary><div class="ref-media"><img src="https://stealthmodels.com/benchmarks/media/9254bfe89a90f7ce.png" alt="Reference for Ground Truth" width="1080" height="1920" loading="lazy" decoding="async"></div></details><div class="pair-slot"><div class="pair"><figure class="plate" data-slot="high" data-result="anthropic-claude-opus-5-5-svg-ground-truth-7c88674d15db6222"><p class="plate-label">Highest score</p><div class="plate-id"><a class="lockup lockup--m" href="https://stealthmodels.com/benchmarks/models/claude-opus-5.5?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="anthropic/claude-opus-5.5"><span class="mount"><img src="/benchmarks/brands/anthropic.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Claude Opus 5.5">Claude Opus 5.5</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="plate-media"><img src="https://stealthmodels.com/benchmarks/media/40fbe76e3262e69e.png" alt="Claude Opus 5.5: Ground Truth benchmark result" width="900" height="1600" loading="lazy" decoding="async"></div><figcaption><p class="plate-score"><span class="ps-label">Benchmark score:</span> <span class="ps-num">80.6</span> <span class="ps-of">out of 100</span></p><p class="plate-meta"><span>Recorded 6 October 2026</span><a class="plain" href="https://stealthmodels.com/benchmarks/models/claude-opus-5.5/ground-truth?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">View result<span class="sr-only"> · Claude Opus 5.5, Ground Truth</span></a></p></figcaption></figure><figure class="plate" data-slot="low" data-result="microsoft-mai-thinking-1-svg-ground-truth-7c88674d15db6222"><p class="plate-label">Lowest score</p><div class="plate-id"><a class="lockup lockup--m" href="https://stealthmodels.com/benchmarks/models/mai-thinking-1?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="microsoft/mai-thinking-1"><span class="mount"><img src="/benchmarks/brands/microsoft-ai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="MAI-Thinking-1">MAI-Thinking-1</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="plate-media"><img src="https://stealthmodels.com/benchmarks/media/ea3bd6787e36fe61.png" alt="MAI-Thinking-1: Ground Truth benchmark result" width="1600" height="900" loading="lazy" decoding="async"></div><figcaption><p class="plate-score"><span class="ps-label">Benchmark score:</span> <span class="ps-num">1.9</span> <span class="ps-of">out of 100</span></p><p class="plate-meta"><span>Recorded 9 October 2026</span><a class="plain" href="https://stealthmodels.com/benchmarks/models/mai-thinking-1/ground-truth?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">View result<span class="sr-only"> · MAI-Thinking-1, Ground Truth</span></a></p></figcaption></figure></div></div><div class="film-row film-row--left" data-film-group="ground-truth" hidden><div class="film-slot"><div class="film film--tall" data-film="7Do7qnXjfic" data-aspect="9:16" data-public-from="2026-10-29T22:00:00Z" hidden><div class="film-copy"><p class="film-label">Physical reasoning: the blueberry riddle</p><p class="film-yt"><a href="https://www.youtube.com/watch?v=7Do7qnXjfic" target="_blank" rel="noopener noreferrer">Watch on YouTube<span class="sr-only"> · Physical reasoning: the blueberry riddle</span></a></p></div><div class="film-frame"><a class="film-poster" href="https://www.youtube.com/watch?v=7Do7qnXjfic" target="_blank" rel="noopener noreferrer" data-film-play aria-label="Watch: Physical reasoning: the blueberry riddle"><img data-src="/assets/films/7Do7qnXjfic.webp" alt="" width="360" height="640" decoding="async"><span class="film-play" aria-hidden="true"></span></a></div></div></div></div><p class="chapter-links"><a class="quiet" href="https://stealthmodels.com/stealthmark?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space#ground-truth" target="_blank" rel="noopener noreferrer">All Ground Truth results <span aria-hidden="true">→</span></a></p></div></article>
<article class="chapter chapter--latch" id="latch" aria-labelledby="latch-title" data-task="latch" data-group="latch-five-levels" data-kind="game"><div class="wrap"><div class="chapter-head"><h2 id="latch-title"><span class="chapter-name">Latch</span><span class="sr-only">: </span><span class="question">Can a model find the best plan, not just a plan, and get better at it level by level?</span></h2><p class="meaning">A high score means the robots coordinate efficiently, avoid irreversible mistakes and improve across levels.</p></div><div class="pair-slot"><div class="pair"><figure class="plate plate--game" data-slot="high"><p class="plate-label">Highest score on the full chain</p><div class="plate-id"><a class="lockup lockup--m" href="https://stealthmodels.com/benchmarks/models/claude-opus-5.5?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="anthropic/claude-opus-5.5"><span class="mount"><img src="/benchmarks/brands/anthropic.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Claude Opus 5.5">Claude Opus 5.5</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="plate-media plate-media--game"><img src="/assets/replay-frames/29aa36983c93b6703c2d.png" alt="Claude Opus 5.5: Latch, level 5" width="600" height="400" loading="lazy" decoding="async"></div><figcaption><p class="plate-score"><span class="ps-label">Benchmark score:</span> <span class="ps-num">87.6</span> <span class="ps-of">out of 100</span></p><p class="plate-cov">5 of 5 levels completed</p><ul class="levels"><li>Level 1 · Cleared · 5.2 points</li><li>Level 2 · Cleared · 10.2 points</li><li>Level 3 · Cleared · 19.2 points</li><li>Level 4 · Cleared · 18.2 points</li><li>Level 5 · Cleared · 20.6 points</li></ul><p class="plate-meta"><a class="btn btn-outline" href="https://stealthmodels.com/benchmarks/models/claude-opus-5.5?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space#game-latch-five-levels" target="_blank" rel="noopener noreferrer">Watch the replay<span class="sr-only"> · Claude Opus 5.5, Latch</span></a></p></figcaption></figure><figure class="plate plate--game" data-slot="low"><p class="plate-label">Lowest score on the full chain</p><div class="plate-id"><a class="lockup lockup--m" href="https://stealthmodels.com/benchmarks/models/glyph-cluster?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="glyph-cluster"><span class="mount"><img src="/_shared/favicons/glyph-cluster.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Glyph Cluster">Glyph Cluster</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="plate-media plate-media--game"><img src="/assets/replay-frames/724b5f8f342cefb1f76c.png" alt="Glyph Cluster: Latch, level 5" width="600" height="400" loading="lazy" decoding="async"></div><figcaption><p class="plate-score"><span class="ps-label">Benchmark score:</span> <span class="ps-num">3.6</span> <span class="ps-of">out of 100</span></p><p class="plate-cov">5 of 5 levels completed</p><ul class="levels"><li>Level 1 · Cleared · 5.2 points</li><li>Level 2 · Out of turns · 2 points</li><li>Level 3 · Finished early · 4.2 points</li><li>Level 4 · Out of turns · 1 points</li><li>Level 5 · Finished early · 9 points</li></ul><p class="plate-meta"><a class="btn btn-outline" href="https://stealthmodels.com/benchmarks/models/glyph-cluster?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space#game-latch-five-levels" target="_blank" rel="noopener noreferrer">Watch the replay<span class="sr-only"> · Glyph Cluster, Latch</span></a></p></figcaption></figure></div></div><div class="film-row film-row--center" data-film-group="latch" hidden><div class="film-slot"><div class="film film--wide" data-film="0tjwIp4lqak" data-aspect="16:9" data-public-from="2026-11-02T16:00:00Z" hidden><div class="film-copy"><p class="film-label">Coordination: two robots in Latch</p><p class="film-yt"><a href="https://www.youtube.com/watch?v=0tjwIp4lqak" target="_blank" rel="noopener noreferrer">Watch on YouTube<span class="sr-only"> · Coordination: two robots in Latch</span></a></p></div><div class="film-frame"><a class="film-poster" href="https://www.youtube.com/watch?v=0tjwIp4lqak" target="_blank" rel="noopener noreferrer" data-film-play aria-label="Watch: Coordination: two robots in Latch"><img data-src="/assets/films/0tjwIp4lqak.webp" alt="" width="800" height="450" decoding="async"><span class="film-play" aria-hidden="true"></span></a></div></div></div></div><p class="chapter-links"><a class="quiet" href="https://stealthmodels.com/stealthmark?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space#latch" target="_blank" rel="noopener noreferrer">All Latch results <span aria-hidden="true">→</span></a></p></div></article>
<article class="chapter chapter--foresight" id="foresight" aria-labelledby="foresight-title" data-task="foresight" data-group="foresight-self-improvement" data-kind="game"><div class="wrap"><div class="chapter-head"><h2 id="foresight-title"><span class="chapter-name">Foresight</span><span class="sr-only">: </span><span class="question">Can a model stay one step ahead of the ghosts?</span></h2><p class="meaning">A high score means the model collects points while anticipating where the ghosts will move.</p></div><div class="pair-slot"><div class="pair"><figure class="plate plate--game" data-slot="high"><p class="plate-label">Highest score on the full chain</p><div class="plate-id"><a class="lockup lockup--m" href="https://stealthmodels.com/benchmarks/models/gpt-6-astra?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="openai/gpt-6-astra"><span class="mount"><img src="/benchmarks/brands/openai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="GPT-6 Astra">GPT-6 Astra</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="plate-media plate-media--game"><img src="/assets/replay-frames/a23e1713ce420337e6ef.png" alt="GPT-6 Astra: Foresight, level 3" width="600" height="400" loading="lazy" decoding="async"></div><figcaption><p class="plate-score"><span class="ps-label">Benchmark score:</span> <span class="ps-num">76.1</span> <span class="ps-of">out of 100</span></p><p class="plate-cov">3 of 3 levels completed</p><ul class="levels"><li>Level 1 · Survived · 27 points</li><li>Level 2 · Survived · 11 points</li><li>Level 3 · Survived · 18 points</li></ul><p class="plate-meta"><a class="btn btn-outline" href="https://stealthmodels.com/benchmarks/models/gpt-6-astra?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space#game-foresight-self-improvement" target="_blank" rel="noopener noreferrer">Watch the replay<span class="sr-only"> · GPT-6 Astra, Foresight</span></a></p></figcaption></figure><figure class="plate plate--game" data-slot="low"><p class="plate-label">Lowest score on the full chain</p><div class="plate-id"><a class="lockup lockup--m" href="https://stealthmodels.com/benchmarks/models/iceman-contributor?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="iceman-contributor"><span class="mount"><img src="/_shared/favicons/iceman-contributor.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Iceman Contributor">Iceman Contributor</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="plate-media plate-media--game"><img src="/assets/replay-frames/76c4266be1ba0b6aceed.png" alt="Iceman Contributor: Foresight, level 3" width="600" height="400" loading="lazy" decoding="async"></div><figcaption><p class="plate-score"><span class="ps-label">Benchmark score:</span> <span class="ps-num">3.2</span> <span class="ps-of">out of 100</span></p><p class="plate-cov">3 of 3 levels completed</p><ul class="levels"><li>Level 1 · Out of lives · 0 points</li><li>Level 2 · Out of lives · 3 points</li><li>Level 3 · Out of lives · 1 points</li></ul><p class="plate-meta"><a class="btn btn-outline" href="https://stealthmodels.com/benchmarks/models/iceman-contributor?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space#game-foresight-self-improvement" target="_blank" rel="noopener noreferrer">Watch the replay<span class="sr-only"> · Iceman Contributor, Foresight</span></a></p></figcaption></figure></div></div><div class="film-row film-row--stage" data-film-group="foresight"><div class="film-slot"><div class="film film--wide" data-film="_cgk-9kjPKQ" data-aspect="16:9"><div class="film-copy"><p class="film-label">Planning ahead: the two traps in Foresight</p><p class="film-yt"><a href="https://www.youtube.com/watch?v=_cgk-9kjPKQ" target="_blank" rel="noopener noreferrer">Watch on YouTube<span class="sr-only"> · Planning ahead: the two traps in Foresight</span></a></p></div><div class="film-frame"><a class="film-poster" href="https://www.youtube.com/watch?v=_cgk-9kjPKQ" target="_blank" rel="noopener noreferrer" data-film-play aria-label="Watch: Planning ahead: the two traps in Foresight"><img src="/assets/films/_cgk-9kjPKQ.webp" loading="lazy" alt="" width="800" height="450" decoding="async"><span class="film-play" aria-hidden="true"></span></a></div></div><div class="film film--tall" data-film="kdG-EBBx0oQ" data-aspect="9:16" data-public-from="2026-11-01T23:00:00Z" hidden><div class="film-copy"><p class="film-label">Planning ahead: three ghosts close in</p><p class="film-yt"><a href="https://www.youtube.com/watch?v=kdG-EBBx0oQ" target="_blank" rel="noopener noreferrer">Watch on YouTube<span class="sr-only"> · Planning ahead: three ghosts close in</span></a></p></div><div class="film-frame"><a class="film-poster" href="https://www.youtube.com/watch?v=kdG-EBBx0oQ" target="_blank" rel="noopener noreferrer" data-film-play aria-label="Watch: Planning ahead: three ghosts close in"><img data-src="/assets/films/kdG-EBBx0oQ.webp" alt="" width="360" height="640" decoding="async"><span class="film-play" aria-hidden="true"></span></a></div></div><div class="film film--tall" data-film="5QABJiwGGxM" data-aspect="9:16" data-public-from="2026-11-08T23:00:00Z" hidden><div class="film-copy"><p class="film-label">Planning on a gaming PC: a local model in Foresight</p><p class="film-yt"><a href="https://www.youtube.com/watch?v=5QABJiwGGxM" target="_blank" rel="noopener noreferrer">Watch on YouTube<span class="sr-only"> · Planning on a gaming PC: a local model in Foresight</span></a></p></div><div class="film-frame"><a class="film-poster" href="https://www.youtube.com/watch?v=5QABJiwGGxM" target="_blank" rel="noopener noreferrer" data-film-play aria-label="Watch: Planning on a gaming PC: a local model in Foresight"><img data-src="/assets/films/5QABJiwGGxM.webp" alt="" width="360" height="640" decoding="async"><span class="film-play" aria-hidden="true"></span></a></div></div></div><div class="film-choices" hidden><p class="film-choices-title">More planning tests</p><ul class="film-choice-list"></ul></div></div><p class="chapter-links"><a class="quiet" href="https://stealthmodels.com/stealthmark?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space#foresight" target="_blank" rel="noopener noreferrer">All Foresight results <span aria-hidden="true">→</span></a></p></div></article>
<article class="chapter chapter--mona-lisa is-portrait" id="mona-lisa" aria-labelledby="mona-lisa-title" data-task="mona-lisa" data-group="mona-lisa" data-kind="svg"><div class="wrap"><div class="chapter-head"><h2 id="mona-lisa-title"><span class="chapter-name">Mona Lisa</span><span class="sr-only">: </span><span class="question">Can a model rebuild what it can describe perfectly?</span></h2><p class="meaning">A high score means the result preserves familiar proportions, expression, composition and light.</p></div><details class="reference"><summary>Reference</summary><div class="ref-media"><img src="https://stealthmodels.com/benchmarks/media/803675074021c0c7.png" alt="Reference for Mona Lisa" width="606" height="791" loading="lazy" decoding="async"></div></details><div class="pair-slot"><div class="pair"><figure class="plate" data-slot="high" data-result="openai-gpt-6-astra-svg-mona-lisa-58920dfd672bf618"><p class="plate-label">Highest score</p><div class="plate-id"><a class="lockup lockup--m" href="https://stealthmodels.com/benchmarks/models/gpt-6-astra?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="openai/gpt-6-astra"><span class="mount"><img src="/benchmarks/brands/openai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="GPT-6 Astra">GPT-6 Astra</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="plate-media"><img src="https://stealthmodels.com/benchmarks/media/45e7730bc430db1d.png" alt="GPT-6 Astra: Mona Lisa benchmark result" width="760" height="1080" loading="lazy" decoding="async"></div><figcaption><p class="plate-score"><span class="ps-label">Benchmark score:</span> <span class="ps-num">66.9</span> <span class="ps-of">out of 100</span></p><p class="plate-meta"><span>Recorded 30 September 2026</span><a class="plain" href="https://stealthmodels.com/benchmarks/models/gpt-6-astra/mona-lisa?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">View result<span class="sr-only"> · GPT-6 Astra, Mona Lisa</span></a></p></figcaption></figure><figure class="plate" data-slot="low" data-result="nvidia-nemotron-3-ultra-550b-a55b-svg-mona-lisa-e7b85d8cd393c26f"><p class="plate-label">Lowest score</p><div class="plate-id"><a class="lockup lockup--m" href="https://stealthmodels.com/benchmarks/models/nemotron-3-ultra-550b-a55b?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="nvidia/nemotron-3-ultra-550b-a55b"><span class="mount"><img src="/benchmarks/brands/nvidia.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Nemotron 3 Ultra">Nemotron 3 Ultra</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="plate-media"><img src="https://stealthmodels.com/benchmarks/media/ad03d1a8517a5e8d.png" alt="Nemotron 3 Ultra: Mona Lisa benchmark result" width="600" height="900" loading="lazy" decoding="async"></div><figcaption><p class="plate-score"><span class="ps-label">Benchmark score:</span> <span class="ps-num">0.5</span> <span class="ps-of">out of 100</span></p><p class="plate-meta"><span>Recorded 4 October 2026</span><a class="plain" href="https://stealthmodels.com/benchmarks/models/nemotron-3-ultra-550b-a55b/mona-lisa?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">View result<span class="sr-only"> · Nemotron 3 Ultra, Mona Lisa</span></a></p></figcaption></figure></div></div><div class="film-row film-row--right" data-film-group="mona-lisa"><div class="film-slot"><div class="film film--tall" data-film="KTz6zx-HIoA" data-aspect="9:16"><div class="film-copy"><p class="film-label">Visual recall: the Mona Lisa from memory</p><p class="film-yt"><a href="https://www.youtube.com/watch?v=KTz6zx-HIoA" target="_blank" rel="noopener noreferrer">Watch on YouTube<span class="sr-only"> · Visual recall: the Mona Lisa from memory</span></a></p></div><div class="film-frame"><a class="film-poster" href="https://www.youtube.com/watch?v=KTz6zx-HIoA" target="_blank" rel="noopener noreferrer" data-film-play aria-label="Watch: Visual recall: the Mona Lisa from memory"><img src="/assets/films/KTz6zx-HIoA.webp" loading="lazy" alt="" width="360" height="640" decoding="async"><span class="film-play" aria-hidden="true"></span></a></div></div></div></div><p class="chapter-links"><a class="quiet" href="https://stealthmodels.com/stealthmark?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space#mona-lisa" target="_blank" rel="noopener noreferrer">All Mona Lisa results <span aria-hidden="true">→</span></a></p></div></article>
<article class="chapter chapter--selfie" id="selfie" aria-labelledby="selfie-title" data-task="selfie" data-group="selfie" data-kind="svg"><div class="wrap"><div class="chapter-head"><h2 id="selfie-title"><span class="chapter-name">Selfie</span><span class="sr-only">: </span><span class="question">Can a model turn a character brief into an original, expressive design?</span></h2><p class="meaning">A high score means the character follows its brief with clear anatomy, expression and materials.</p></div><div class="pair-slot"><p class="pair-note">Different character briefs</p><div class="pair"><figure class="plate" data-slot="high" data-result="openai-gpt-6-astra-svg-selfie-efa9c627eda55ad5"><p class="plate-label">Highest score</p><div class="plate-id"><a class="lockup lockup--m" href="https://stealthmodels.com/benchmarks/models/gpt-6-astra?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="openai/gpt-6-astra"><span class="mount"><img src="/benchmarks/brands/openai.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="GPT-6 Astra">GPT-6 Astra</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="plate-media"><img src="https://stealthmodels.com/benchmarks/media/5727401e8bbcdd5e.png" alt="GPT-6 Astra: Selfie benchmark result" width="1200" height="1200" loading="lazy" decoding="async"></div><figcaption><p class="plate-score"><span class="ps-label">Benchmark score:</span> <span class="ps-num">80.9</span> <span class="ps-of">out of 100</span></p><p class="plate-meta"><span>Recorded 30 September 2026</span><a class="plain" href="https://stealthmodels.com/benchmarks/models/gpt-6-astra/selfie?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">View result<span class="sr-only"> · GPT-6 Astra, Selfie</span></a></p></figcaption></figure><figure class="plate" data-slot="low" data-result="aleph-alpha-kolibri-1-svg-selfie-e5bcbc64de37f34f"><p class="plate-label">Lowest score</p><div class="plate-id"><a class="lockup lockup--m" href="https://stealthmodels.com/benchmarks/models/kolibri-1?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="Aleph-Alpha/Kolibri-1"><span class="mount"><img src="/benchmarks/brands/aleph-alpha.svg" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Kolibri-1">Kolibri-1</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="plate-media"><img src="https://stealthmodels.com/benchmarks/media/fba0e930cae81873.png" alt="Kolibri-1: Selfie benchmark result" width="400" height="500" loading="lazy" decoding="async"></div><figcaption><p class="plate-score"><span class="ps-label">Benchmark score:</span> <span class="ps-num">22.9</span> <span class="ps-of">out of 100</span></p><p class="plate-meta"><span>Recorded 5 October 2026</span><a class="plain" href="https://stealthmodels.com/benchmarks/models/kolibri-1/selfie?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">View result<span class="sr-only"> · Kolibri-1, Selfie</span></a></p></figcaption></figure></div></div><p class="chapter-links"><a class="quiet" href="https://stealthmodels.com/stealthmark?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space#selfie" target="_blank" rel="noopener noreferrer">All Selfie results <span aria-hidden="true">→</span></a></p></div></article>
<article class="chapter chapter--beach" id="beach" aria-labelledby="beach-title" data-task="beach" data-group="beach" data-kind="svg"><div class="wrap"><div class="chapter-head"><h2 id="beach-title"><span class="chapter-name">Beach</span><span class="sr-only">: </span><span class="question">Can a model compose a whole scene that is both correct and beautiful?</span></h2><p class="meaning">A high score means perspective, scale, reflections and composition work together across the scene.</p></div><div class="pair-slot"><div class="pair"><figure class="plate" data-slot="high" data-result="anthropic-claude-sonnet-5-5-svg-beach-d030aa6f45fe19e9"><p class="plate-label">Highest score</p><div class="plate-id"><a class="lockup lockup--m" href="https://stealthmodels.com/benchmarks/models/claude-sonnet-5.5?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="anthropic/claude-sonnet-5.5"><span class="mount"><img src="/benchmarks/brands/anthropic.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Claude Sonnet 5.5">Claude Sonnet 5.5</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="plate-media"><img src="https://stealthmodels.com/benchmarks/media/98f3acd10bb91df6.png" alt="Claude Sonnet 5.5: Beach benchmark result" width="1200" height="800" loading="lazy" decoding="async"></div><figcaption><p class="plate-score"><span class="ps-label">Benchmark score:</span> <span class="ps-num">87.0</span> <span class="ps-of">out of 100</span></p><p class="plate-meta"><span>Recorded 30 September 2026</span><a class="plain" href="https://stealthmodels.com/benchmarks/models/claude-sonnet-5.5/beach?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">View result<span class="sr-only"> · Claude Sonnet 5.5, Beach</span></a></p></figcaption></figure><figure class="plate" data-slot="low" data-result="qwen-qwen3-5-9b-svg-beach-d030aa6f45fe19e9"><p class="plate-label">Lowest score</p><div class="plate-id"><a class="lockup lockup--m" href="https://stealthmodels.com/benchmarks/models/qwen3.5-9b?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="qwen/qwen3.5-9b"><span class="mount"><img src="/benchmarks/brands/qwen.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Qwen3.5-9B">Qwen3.5-9B</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="plate-media"><img src="https://stealthmodels.com/benchmarks/media/f6ca1994f9d37e65.png" alt="Qwen3.5-9B: Beach benchmark result" width="1200" height="800" loading="lazy" decoding="async"></div><figcaption><p class="plate-score"><span class="ps-label">Benchmark score:</span> <span class="ps-num">12.8</span> <span class="ps-of">out of 100</span></p><p class="plate-meta"><span>Recorded 5 October 2026</span><a class="plain" href="https://stealthmodels.com/benchmarks/models/qwen3.5-9b/beach?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">View result<span class="sr-only"> · Qwen3.5-9B, Beach</span></a></p></figcaption></figure></div></div><p class="chapter-links"><a class="quiet" href="https://stealthmodels.com/stealthmark?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space#beach" target="_blank" rel="noopener noreferrer">All Beach results <span aria-hidden="true">→</span></a></p></div></article>
<article class="chapter chapter--raccoon" id="raccoon" aria-labelledby="raccoon-title" data-task="raccoon" data-group="raccoon" data-kind="svg"><div class="wrap"><div class="chapter-head"><h2 id="raccoon-title"><span class="chapter-name">Raccoon</span><span class="sr-only">: </span><span class="question">Can a model show characters actually doing something together?</span></h2><p class="meaning">A high score means the dog, raccoon and jet ski connect through believable contact, perspective and motion.</p></div><div class="pair-slot"><div class="pair"><figure class="plate" data-slot="high" data-result="google-gemini-4-argon-svg-raccoon-b4cedf42a4a26a4b"><p class="plate-label">Highest score</p><div class="plate-id"><a class="lockup lockup--m" href="https://stealthmodels.com/benchmarks/models/gemini-4-argon?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="google/gemini-4-argon"><span class="mount"><img src="/benchmarks/brands/google.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Gemini 4 Argon">Gemini 4 Argon</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="plate-media"><img src="https://stealthmodels.com/benchmarks/media/f5af6a339a32556b.png" alt="Gemini 4 Argon: Raccoon benchmark result" width="1200" height="800" loading="lazy" decoding="async"></div><figcaption><p class="plate-score"><span class="ps-label">Benchmark score:</span> <span class="ps-num">81.3</span> <span class="ps-of">out of 100</span></p><p class="plate-meta"><span>Recorded 13 October 2026</span><a class="plain" href="https://stealthmodels.com/benchmarks/models/gemini-4-argon/raccoon?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">View result<span class="sr-only"> · Gemini 4 Argon, Raccoon</span></a></p></figcaption></figure><figure class="plate" data-slot="low" data-result="aleph-alpha-kolibri-1-svg-raccoon-b4cedf42a4a26a4b"><p class="plate-label">Lowest score</p><div class="plate-id"><a class="lockup lockup--m" href="https://stealthmodels.com/benchmarks/models/kolibri-1?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="Aleph-Alpha/Kolibri-1"><span class="mount"><img src="/benchmarks/brands/aleph-alpha.svg" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Kolibri-1">Kolibri-1</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a></div><div class="plate-media"><img src="https://stealthmodels.com/benchmarks/media/e005d9c8a607783d.png" alt="Kolibri-1: Raccoon benchmark result" width="800" height="600" loading="lazy" decoding="async"></div><figcaption><p class="plate-score"><span class="ps-label">Benchmark score:</span> <span class="ps-num">10.0</span> <span class="ps-of">out of 100</span></p><p class="plate-meta"><span>Recorded 5 October 2026</span><a class="plain" href="https://stealthmodels.com/benchmarks/models/kolibri-1/raccoon?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">View result<span class="sr-only"> · Kolibri-1, Raccoon</span></a></p></figcaption></figure></div></div><div class="film-row film-row--left" data-film-group="raccoon" hidden><div class="film-slot"><div class="film film--tall" data-film="oc0uWGJ8-Zg" data-aspect="9:16" data-public-from="2026-11-12T23:00:00Z" hidden><div class="film-copy"><p class="film-label">Composition: characters that interact</p><p class="film-yt"><a href="https://www.youtube.com/watch?v=oc0uWGJ8-Zg" target="_blank" rel="noopener noreferrer">Watch on YouTube<span class="sr-only"> · Composition: characters that interact</span></a></p></div><div class="film-frame"><a class="film-poster" href="https://www.youtube.com/watch?v=oc0uWGJ8-Zg" target="_blank" rel="noopener noreferrer" data-film-play aria-label="Watch: Composition: characters that interact"><img data-src="/assets/films/oc0uWGJ8-Zg.webp" alt="" width="360" height="640" decoding="async"><span class="film-play" aria-hidden="true"></span></a></div></div></div></div><p class="chapter-links"><a class="quiet" href="https://stealthmodels.com/stealthmark?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space#raccoon" target="_blank" rel="noopener noreferrer">All Raccoon results <span aria-hidden="true">→</span></a></p></div></article>
</section>
<section class="play" id="play" aria-labelledby="play-title">
<div class="wrap">
<h2 id="play-title">Play it yourself</h2>
<p class="play-line">The same levels and rules the models get. See where your score would place.</p>
<div class="play-grid"><div class="play-entry"><img src="/assets/boards/game-preview-foresight-self-improvement.png" alt="Foresight benchmark board" width="600" height="400" loading="lazy" decoding="async"><h3>Foresight</h3><p>Can you stay one step ahead of the ghosts?</p><a class="btn btn-gold" href="https://stealthmodels.com/play/foresight?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">Play Foresight</a></div><div class="play-entry"><img src="/assets/boards/game-preview-latch-five-levels.png" alt="Latch benchmark board" width="600" height="400" loading="lazy" decoding="async"><h3>Latch</h3><p>Can you steer two robots without one careless push?</p><a class="btn btn-gold" href="https://stealthmodels.com/play/latch?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">Play Latch</a></div></div>
</div>
</section>
<section class="stealth" id="stealth" aria-labelledby="stealth-title">
<div class="wrap">
<h2 id="stealth-title">Still in stealth</h2>
<p class="section-lead">No company has claimed these models yet. Follow us to see who made them.</p>
<div class="stealth-layout">
<div class="stealth-grid"><div class="stealth-entry" data-stealth="iceman-contributor"><img src="/assets/mascots/iceman-contributor-face.webp" alt="Iceman Contributor mascot" loading="lazy" decoding="async"><div class="stealth-id"><a class="lockup lockup--xl" href="https://stealthmodels.com/benchmarks/models/iceman-contributor?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="iceman-contributor"><span class="mount"><img src="/_shared/favicons/iceman-contributor.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Iceman Contributor">Iceman Contributor</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a><p><a class="sec" href="https://huggingface.co/stealth-model/iceman-contributor" target="_blank" rel="noopener noreferrer">Our findings on Hugging Face<span class="sr-only"> · Iceman Contributor</span></a></p></div></div><div class="stealth-entry" data-stealth="glyph-cluster"><img src="/assets/mascots/cluster-v2.svg" alt="Glyph Cluster mascots" loading="lazy" decoding="async"><div class="stealth-id"><a class="lockup lockup--xl" href="https://stealthmodels.com/benchmarks/models/glyph-cluster?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer" data-model="glyph-cluster"><span class="mount"><img src="/_shared/favicons/glyph-cluster.png" alt="" loading="lazy" decoding="async" width="40" height="40"></span><span class="lname" title="Glyph Cluster">Glyph Cluster</span><svg class="arrow" viewBox="0 0 14 14" width="14" height="14" aria-hidden="true"><path d="M4 10 10 4M5 4h5v5" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/></svg></a><p><a class="sec" href="https://huggingface.co/stealth-model/glyph-cluster" target="_blank" rel="noopener noreferrer">Our findings on Hugging Face<span class="sr-only"> · Glyph Cluster</span></a></p></div></div><p class="follow"><a class="btn btn-outline" href="https://huggingface.co/stealth-model" target="_blank" rel="noopener noreferrer">Follow Stealth Models on Hugging Face <span aria-hidden="true">→</span></a></p></div>
<div class="film-row film-row--stealth" data-film-group="stealth"><div class="film-slot"><div class="film film--tall" data-film="l5hdM8lsPfI" data-aspect="9:16"><div class="film-copy"><p class="film-label">Iceman Contributor: its results and our clues so far</p><p class="film-yt"><a href="https://www.youtube.com/watch?v=l5hdM8lsPfI" target="_blank" rel="noopener noreferrer">Watch on YouTube<span class="sr-only"> · Iceman Contributor: its results and our clues so far</span></a></p></div><div class="film-frame"><a class="film-poster" href="https://www.youtube.com/watch?v=l5hdM8lsPfI" target="_blank" rel="noopener noreferrer" data-film-play aria-label="Watch: Iceman Contributor: its results and our clues so far"><img src="/assets/films/l5hdM8lsPfI.webp" loading="lazy" alt="" width="360" height="640" decoding="async"><span class="film-play" aria-hidden="true"></span></a></div></div></div></div>
</div>
</div>
</section>
</main>
<footer class="cite" id="cite">
<div class="wrap cite-grid">
<div>
<h2>Cite StealthMark</h2>
<p class="cite-text" id="cite-text" data-cite-text>Stealth Models (2026). StealthMark Intelligence Benchmarks, version 1.8.8. https://stealthmodels.com/stealthmark</p>
<p class="cite-actions"><button type="button" class="btn btn-outline" data-copy="cite-text" data-done="Citation copied">Copy citation</button><button type="button" class="btn btn-outline" data-copy="cite-bibtex" data-done="BibTeX copied">Copy BibTeX</button><span class="copy-status" role="status" aria-live="polite" data-fail="Copy unavailable. Select the text to copy it."></span></p>
<details class="bib"><summary>BibTeX</summary><pre class="cite-bibtex" id="cite-bibtex" tabindex="0"><code data-cite-bibtex>@misc{stealthmark2026,
author = {{Stealth Models}},
title = {{StealthMark} Intelligence Benchmarks},
year = {2026},
version = {1.8.8},
url = {https://stealthmodels.com/stealthmark}
}</code></pre></details>
</div>
<ul class="cite-links">
<li><a class="quiet" href="https://stealthmodels.com/stealthmark?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">How StealthMark works <span aria-hidden="true">→</span></a></li>
<li><a class="quiet" href="https://stealthmodels.com/benchmarks?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">Full benchmark results <span aria-hidden="true">→</span></a></li>
<li><a class="quiet" href="https://huggingface.co/stealth-model" target="_blank" rel="noopener noreferrer">Stealth Models on Hugging Face <span aria-hidden="true">→</span></a></li>
<li><a class="quiet" href="https://stealthmodels.com/image-license?utm_source=huggingface&amp;utm_medium=referral&amp;utm_campaign=stealthmark-space" target="_blank" rel="noopener noreferrer">Image licence <span aria-hidden="true">→</span></a> (CC BY 4.0)</li>
<li><a class="quiet" href="https://huggingface.co/spaces/stealth-model/stealthmark" target="_blank" rel="noopener noreferrer">Like this Space on Hugging Face <span aria-hidden="true">→</span></a></li>
</ul>
</div>
</footer>
<dialog class="picker" id="picker" aria-labelledby="picker-title">
<div class="pk-head"><h2 id="picker-title"></h2><button type="button" class="pk-close" data-pk-close>Close</button></div>
<input id="pk-search" class="pk-search" type="search" placeholder="Search models" aria-label="Search models" autocomplete="off" spellcheck="false">
<ul class="pk-list" id="pk-list"></ul>
<p class="pk-empty" hidden>No models match these filters.</p>
</dialog>
</body>
</html>