Update README.md
Browse files
README.md
CHANGED
|
@@ -36,6 +36,8 @@ datasets:
|
|
| 36 |
- OpenSPG/KAG-Thinker-training-dataset
|
| 37 |
- Gryphe/ChatGPT-4o-Writing-Prompts
|
| 38 |
library_name: transformers
|
|
|
|
|
|
|
| 39 |
---
|
| 40 |
|
| 41 |
<div align="center" style="font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;">
|
|
@@ -45,7 +47,7 @@ library_name: transformers
|
|
| 45 |

|
| 46 |
|
| 47 |
|
| 48 |
-
<h1 style="color: #0ea5e9; font-weight: 800; font-size: 2.8em; margin-bottom: 5px; letter-spacing: -1px;">
|
| 49 |
<h3 style="color: #64748b; font-weight: 400; margin-top: 0; font-size: 1.2em;"><i>Türkiye’s Fastest Lightweight Multimodal & Reasoning AI</i></h3>
|
| 50 |
|
| 51 |
<p style="margin-top: 15px;">
|
|
@@ -69,8 +71,8 @@ While large models dominate cloud servers, Next2-Air is designed to bring top-ti
|
|
| 69 |
|
| 70 |
## ⚡ Highlights
|
| 71 |
|
| 72 |
-
<div style="background:
|
| 73 |
-
<ul style="margin: 0; padding-left: 20px; line-height: 1.6; color: #
|
| 74 |
<li>🇹🇷 <strong>Perfected in Türkiye:</strong> Fine-tuned with cultural nuance, ensuring natural, fluent, and highly accurate Turkish responses.</li>
|
| 75 |
<li>💨 <strong>"Air" Speed & Efficiency:</strong> Only 2 Billion parameters. Runs blazingly fast on MacBooks, mid-range PCs, and edge hardware without needing massive GPUs.</li>
|
| 76 |
<li>🧠 <strong>Native Thinking Mode:</strong> Despite its small size, it leverages Chain-of-Thought (<code><think></code>) to logically deduce answers before speaking.</li>
|
|
@@ -87,19 +89,19 @@ Next2-Air (2B) redefines what is possible in the ultra-lightweight category. Thr
|
|
| 87 |
|
| 88 |
### 📝 Text, Reasoning & Instruction Following
|
| 89 |
|
| 90 |
-
<div style="overflow-x: auto; box-shadow: 0 4px 6px rgba(0,0,0,0.05); border-radius:
|
| 91 |
-
<table style="width: 100%; border-collapse: collapse; text-align: center; font-family: sans-serif; background: #
|
| 92 |
<thead>
|
| 93 |
-
<tr style="background-color: #
|
| 94 |
-
<th style="padding: 14px; text-align: left; padding-left: 20px; border-radius:
|
| 95 |
-
<th style="padding: 14px; font-size: 1.1em;">Next2-Air (2B)
|
| 96 |
<th style="padding: 14px;">Qwen 3.5 (2B)</th>
|
| 97 |
<th style="padding: 14px;">Gemma-2 (2B)</th>
|
| 98 |
-
<th style="padding: 14px; border-radius: 0
|
| 99 |
</tr>
|
| 100 |
</thead>
|
| 101 |
-
<tbody style="color: #
|
| 102 |
-
<tr style="border-bottom: 1px solid #f1f5f9; background-color: #
|
| 103 |
<td style="padding: 12px; text-align: left; padding-left: 20px; color: #0284c7;">MMLU-Pro (Thinking)</td>
|
| 104 |
<td style="padding: 12px; color: #0ea5e9;">68.2%</td>
|
| 105 |
<td style="padding: 12px;">66.5%</td>
|
|
@@ -107,13 +109,13 @@ Next2-Air (2B) redefines what is possible in the ultra-lightweight category. Thr
|
|
| 107 |
<td style="padding: 12px;">68.4%</td>
|
| 108 |
</tr>
|
| 109 |
<tr style="border-bottom: 1px solid #f1f5f9;">
|
| 110 |
-
<td style="padding: 12px; text-align: left; padding-left: 20px;">MMLU-Redux</td>
|
| 111 |
<td style="padding: 12px; font-weight: bold; color: #0ea5e9;">82.1%</td>
|
| 112 |
<td style="padding: 12px;">79.6%</td>
|
| 113 |
<td style="padding: 12px;">75.3%</td>
|
| 114 |
<td style="padding: 12px;">79.5%</td>
|
| 115 |
</tr>
|
| 116 |
-
<tr style="border-bottom: 1px solid #f1f5f9; background-color: #
|
| 117 |
<td style="padding: 12px; text-align: left; padding-left: 20px; color: #0284c7;">IFEval (Instruction)</td>
|
| 118 |
<td style="padding: 12px; color: #0ea5e9;">82.5%</td>
|
| 119 |
<td style="padding: 12px;">78.6%</td>
|
|
@@ -121,7 +123,7 @@ Next2-Air (2B) redefines what is possible in the ultra-lightweight category. Thr
|
|
| 121 |
<td style="padding: 12px;">77.4%</td>
|
| 122 |
</tr>
|
| 123 |
<tr style="border-bottom: 1px solid #f1f5f9;">
|
| 124 |
-
<td style="padding: 12px; text-align: left; padding-left: 20px;">TAU2-Bench (Agent)</td>
|
| 125 |
<td style="padding: 12px; font-weight: bold; color: #0ea5e9;">52.4%</td>
|
| 126 |
<td style="padding: 12px;">48.8%</td>
|
| 127 |
<td style="padding: 12px;">--</td>
|
|
@@ -135,33 +137,33 @@ Next2-Air (2B) redefines what is possible in the ultra-lightweight category. Thr
|
|
| 135 |
|
| 136 |
Next2-Air features a highly capable visual encoder, allowing it to process spatial intelligence, OCR, and document understanding tasks efficiently.
|
| 137 |
|
| 138 |
-
<div style="overflow-x: auto; box-shadow: 0 4px 6px rgba(0,0,0,0.05); border-radius: 8px; margin-top: 15px;">
|
| 139 |
-
<table style="width: 100%; border-collapse: collapse; text-align: center; font-family: sans-serif; background: #
|
| 140 |
<thead>
|
| 141 |
-
<tr style="background-color: #
|
| 142 |
-
<th style="padding: 14px; text-align: left; padding-left: 20px; border-radius:
|
| 143 |
-
<th style="padding: 14px; font-size: 1.1em;">Next2-Air (2B)
|
| 144 |
-
<th style="padding: 14px; border-radius: 0
|
| 145 |
</tr>
|
| 146 |
</thead>
|
| 147 |
-
<tbody style="color: #
|
| 148 |
<tr style="border-bottom: 1px solid #f1f5f9;">
|
| 149 |
-
<td style="padding: 12px; text-align: left; padding-left: 20px;">MMMU (General VQA)</td>
|
| 150 |
<td style="padding: 12px; font-weight: bold; color: #0ea5e9;">66.5%</td>
|
| 151 |
<td style="padding: 12px;">64.2%</td>
|
| 152 |
</tr>
|
| 153 |
-
<tr style="border-bottom: 1px solid #f1f5f9; background-color: #
|
| 154 |
-
<td style="padding: 12px; text-align: left; padding-left: 20px;">MathVision</td>
|
| 155 |
<td style="padding: 12px; font-weight: bold; color: #0ea5e9;">78.1%</td>
|
| 156 |
<td style="padding: 12px;">76.7%</td>
|
| 157 |
</tr>
|
| 158 |
<tr style="border-bottom: 1px solid #f1f5f9;">
|
| 159 |
-
<td style="padding: 12px; text-align: left; padding-left: 20px;">OCRBench</td>
|
| 160 |
<td style="padding: 12px; font-weight: bold; color: #0ea5e9;">86.0%</td>
|
| 161 |
<td style="padding: 12px;">84.5%</td>
|
| 162 |
</tr>
|
| 163 |
-
<tr style="border-bottom: 1px solid #f1f5f9; background-color: #
|
| 164 |
-
<td style="padding: 12px; text-align: left; padding-left: 20px;">VideoMME (w/ sub)</td>
|
| 165 |
<td style="padding: 12px; font-weight: bold; color: #0ea5e9;">77.8%</td>
|
| 166 |
<td style="padding: 12px;">75.6%</td>
|
| 167 |
</tr>
|
|
@@ -256,8 +258,8 @@ Next2-Air is released under the **Apache 2.0 License**. We strongly believe in e
|
|
| 256 |
|
| 257 |
---
|
| 258 |
|
| 259 |
-
<div align="center" style="margin-top: 40px; padding: 25px; border-top: 1px solid #e0f2fe; background: #
|
| 260 |
-
<p style="color: #
|
| 261 |
<strong>Next2-Air</strong> — Hafif, Hızlı, Akıllı. Uç cihazlardan buluta, Türkiye'nin yeni nesil çevik yapay zekası. 🌬️
|
| 262 |
</p>
|
| 263 |
</div>
|
|
|
|
| 36 |
- OpenSPG/KAG-Thinker-training-dataset
|
| 37 |
- Gryphe/ChatGPT-4o-Writing-Prompts
|
| 38 |
library_name: transformers
|
| 39 |
+
base_model:
|
| 40 |
+
- thelamapi/next2-air
|
| 41 |
---
|
| 42 |
|
| 43 |
<div align="center" style="font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;">
|
|
|
|
| 47 |

|
| 48 |
|
| 49 |
|
| 50 |
+
<h1 style="color: #0ea5e9; font-weight: 800; font-size: 2.8em; margin-bottom: 5px; letter-spacing: -1px;">Next2-Air (2B)</h1>
|
| 51 |
<h3 style="color: #64748b; font-weight: 400; margin-top: 0; font-size: 1.2em;"><i>Türkiye’s Fastest Lightweight Multimodal & Reasoning AI</i></h3>
|
| 52 |
|
| 53 |
<p style="margin-top: 15px;">
|
|
|
|
| 71 |
|
| 72 |
## ⚡ Highlights
|
| 73 |
|
| 74 |
+
<div style="background: #232323; border-left: 5px solid #0ea5e9; padding: 20px; width:fit-content; border-radius: 16px; font-family: sans-serif;">
|
| 75 |
+
<ul style="margin: 0; padding-left: 20px; line-height: 1.6; color: #808080;">
|
| 76 |
<li>🇹🇷 <strong>Perfected in Türkiye:</strong> Fine-tuned with cultural nuance, ensuring natural, fluent, and highly accurate Turkish responses.</li>
|
| 77 |
<li>💨 <strong>"Air" Speed & Efficiency:</strong> Only 2 Billion parameters. Runs blazingly fast on MacBooks, mid-range PCs, and edge hardware without needing massive GPUs.</li>
|
| 78 |
<li>🧠 <strong>Native Thinking Mode:</strong> Despite its small size, it leverages Chain-of-Thought (<code><think></code>) to logically deduce answers before speaking.</li>
|
|
|
|
| 89 |
|
| 90 |
### 📝 Text, Reasoning & Instruction Following
|
| 91 |
|
| 92 |
+
<div style="overflow-x: auto; box-shadow: 0 4px 6px rgba(0,0,0,0.05); width:fit-content; border-radius: 16px;">
|
| 93 |
+
<table style="width: 100%; border-collapse: collapse; text-align: center; font-family: sans-serif; background: #232323; min-width: 800px;">
|
| 94 |
<thead>
|
| 95 |
+
<tr style="background-color: #232323; color: white;">
|
| 96 |
+
<th style="padding: 14px; text-align: left; padding-left: 20px; border-radius: 16px 0 0 0;">Benchmark</th>
|
| 97 |
+
<th style="padding: 14px; font-size: 1.1em;">Next2-Air (2B)</th>
|
| 98 |
<th style="padding: 14px;">Qwen 3.5 (2B)</th>
|
| 99 |
<th style="padding: 14px;">Gemma-2 (2B)</th>
|
| 100 |
+
<th style="padding: 14px; border-radius: 0 16px 0 0;">Llama-3.2 (3B)</th>
|
| 101 |
</tr>
|
| 102 |
</thead>
|
| 103 |
+
<tbody style="color: #808080;">
|
| 104 |
+
<tr style="border-bottom: 1px solid #f1f5f9; background-color: #232323; font-weight: 600;">
|
| 105 |
<td style="padding: 12px; text-align: left; padding-left: 20px; color: #0284c7;">MMLU-Pro (Thinking)</td>
|
| 106 |
<td style="padding: 12px; color: #0ea5e9;">68.2%</td>
|
| 107 |
<td style="padding: 12px;">66.5%</td>
|
|
|
|
| 109 |
<td style="padding: 12px;">68.4%</td>
|
| 110 |
</tr>
|
| 111 |
<tr style="border-bottom: 1px solid #f1f5f9;">
|
| 112 |
+
<td style="padding: 12px; text-align: left; padding-left: 20px; color: #0284c7;;">MMLU-Redux</td>
|
| 113 |
<td style="padding: 12px; font-weight: bold; color: #0ea5e9;">82.1%</td>
|
| 114 |
<td style="padding: 12px;">79.6%</td>
|
| 115 |
<td style="padding: 12px;">75.3%</td>
|
| 116 |
<td style="padding: 12px;">79.5%</td>
|
| 117 |
</tr>
|
| 118 |
+
<tr style="border-bottom: 1px solid #f1f5f9; background-color: #232323; font-weight: 600;">
|
| 119 |
<td style="padding: 12px; text-align: left; padding-left: 20px; color: #0284c7;">IFEval (Instruction)</td>
|
| 120 |
<td style="padding: 12px; color: #0ea5e9;">82.5%</td>
|
| 121 |
<td style="padding: 12px;">78.6%</td>
|
|
|
|
| 123 |
<td style="padding: 12px;">77.4%</td>
|
| 124 |
</tr>
|
| 125 |
<tr style="border-bottom: 1px solid #f1f5f9;">
|
| 126 |
+
<td style="padding: 12px; text-align: left; padding-left: 20px; color: #0284c7;">TAU2-Bench (Agent)</td>
|
| 127 |
<td style="padding: 12px; font-weight: bold; color: #0ea5e9;">52.4%</td>
|
| 128 |
<td style="padding: 12px;">48.8%</td>
|
| 129 |
<td style="padding: 12px;">--</td>
|
|
|
|
| 137 |
|
| 138 |
Next2-Air features a highly capable visual encoder, allowing it to process spatial intelligence, OCR, and document understanding tasks efficiently.
|
| 139 |
|
| 140 |
+
<div style="overflow-x: auto; box-shadow: 0 4px 6px rgba(0,0,0,0.05); border-radius: 8px; margin-top: 15px;width:fit-content; ">
|
| 141 |
+
<table style="width: 100%; border-collapse: collapse; text-align: center; font-family: sans-serif; background: #232323; min-width: 800px;">
|
| 142 |
<thead>
|
| 143 |
+
<tr style="background-color: #232323; color: white;">
|
| 144 |
+
<th style="padding: 14px; text-align: left; padding-left: 20px; border-radius: 16px 0 0 0;">Benchmark</th>
|
| 145 |
+
<th style="padding: 14px; font-size: 1.1em;">Next2-Air (2B)</th>
|
| 146 |
+
<th style="padding: 14px; border-radius: 0 16px 0 0;">Base Qwen3.5-2B</th>
|
| 147 |
</tr>
|
| 148 |
</thead>
|
| 149 |
+
<tbody style="color: #808080;">
|
| 150 |
<tr style="border-bottom: 1px solid #f1f5f9;">
|
| 151 |
+
<td style="padding: 12px; text-align: left; padding-left: 20px; color: #0284c7;;">MMMU (General VQA)</td>
|
| 152 |
<td style="padding: 12px; font-weight: bold; color: #0ea5e9;">66.5%</td>
|
| 153 |
<td style="padding: 12px;">64.2%</td>
|
| 154 |
</tr>
|
| 155 |
+
<tr style="border-bottom: 1px solid #f1f5f9; background-color: #232323;">
|
| 156 |
+
<td style="padding: 12px; text-align: left; padding-left: 20px; color: #0284c7;">MathVision</td>
|
| 157 |
<td style="padding: 12px; font-weight: bold; color: #0ea5e9;">78.1%</td>
|
| 158 |
<td style="padding: 12px;">76.7%</td>
|
| 159 |
</tr>
|
| 160 |
<tr style="border-bottom: 1px solid #f1f5f9;">
|
| 161 |
+
<td style="padding: 12px; text-align: left; padding-left: 20px; color: #0284c7;">OCRBench</td>
|
| 162 |
<td style="padding: 12px; font-weight: bold; color: #0ea5e9;">86.0%</td>
|
| 163 |
<td style="padding: 12px;">84.5%</td>
|
| 164 |
</tr>
|
| 165 |
+
<tr style="border-bottom: 1px solid #f1f5f9; background-color: #232323;">
|
| 166 |
+
<td style="padding: 12px; text-align: left; padding-left: 20px; color: #0284c7;">VideoMME (w/ sub)</td>
|
| 167 |
<td style="padding: 12px; font-weight: bold; color: #0ea5e9;">77.8%</td>
|
| 168 |
<td style="padding: 12px;">75.6%</td>
|
| 169 |
</tr>
|
|
|
|
| 258 |
|
| 259 |
---
|
| 260 |
|
| 261 |
+
<div align="center" style="margin-top: 40px; padding: 25px; border-top: 1px solid #e0f2fe; background: #232323; border-radius: 8px;width:fit-content; ">
|
| 262 |
+
<p style="color: #808080; font-size: 15px; margin: 0;">
|
| 263 |
<strong>Next2-Air</strong> — Hafif, Hızlı, Akıllı. Uç cihazlardan buluta, Türkiye'nin yeni nesil çevik yapay zekası. 🌬️
|
| 264 |
</p>
|
| 265 |
</div>
|