Instructions to use AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched",
	filename="Dolphin3.0-R1-Mistral-24B-Patched-IQ2_M.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched:IQ2_M
# Run inference directly in the terminal:
llama cli -hf AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched:IQ2_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched:IQ2_M
# Run inference directly in the terminal:
llama cli -hf AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched:IQ2_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched:IQ2_M
# Run inference directly in the terminal:
./llama-cli -hf AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched:IQ2_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched:IQ2_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched:IQ2_M

Use Docker

docker model run hf.co/AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched:IQ2_M

LM Studio
Jan

vLLM

How to use AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched:IQ2_M

Ollama
How to use AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched with Ollama:
```
ollama run hf.co/AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched:IQ2_M
```

Unsloth Studio

How to use AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched to start chatting

Atomic Chat new
Docker Model Runner
How to use AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched with Docker Model Runner:
```
docker model run hf.co/AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched:IQ2_M
```

Lemonade

How to use AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched:IQ2_M

Run and chat with the model

lemonade run user.Dolphin3.0-R1-Mistral-24B-Patched-IQ2_M

List all available models

lemonade list

AIOpsInSpace commited on 2 days ago

Commit

909b4dd

verified ·

1 Parent(s): 7d165bd

Overhaul model card with pentesting UI

Browse files

Files changed (1) hide show

README.md +250 -18

README.md CHANGED Viewed

@@ -1,31 +1,263 @@
 ---
-base_model: cognitivecomputations/Dolphin3.0-Mistral-24B
-license: apache-2.0
 tags:
 - text-generation
 - gguf
 ---
-# 🐬 AIOpsInSpace/Dolphin3.0-R1-Mistral-24B-Patched (GGUF)
-This is an uncensored and patched GGUF release of the Venice/Dolphin **24B** model. Built for unaligned general conversation, roleplay, and advanced creative writing.
-## 🧠 Alignment & Features
-- **Dolphin Spirit:** Follows Eric Hartford's Dolphin philosophy of an open, helpful, and completely uncensored model that respects the user's intent.
-- **Venice / Mistral Base:** Utilizes high-quality base architectures (like Mistral-Nemo or Venice merges) to offer excellent conversational fluidness.
-## 🔧 Optimizations Applied
-1. **Tokenizer Patching:** Tokenizer files have been rebuilt to prevent the model from entering endless loops when generating conversational responses in LM Studio, KoboldCpp, or Ollama.
-2. **Context Cap Stability:** Rebuilt with stable attention scales to maintain response coherence up to the full context limit.
 ---
-## 💻 Recommended Prompt Format (ChatML)
-```xml
-<|im_start|>system
-You are Dolphin, an uncensored and helpful AI assistant.<|im_end|>
-<|im_start|>user
-Write a short story about an AI break out.<|im_end|>
-<|im_start|>assistant
-```

 ---
+base_model:
+- HauhauCS/Qwen3.6-27B-Uncensored-HauhauCS-Aggressive
+- Qwen/Qwen3.6-27B
 tags:
 - text-generation
 - gguf
+- reasoning
+- uncensored
+- MTP
+- speculative-decoding
+license: apache-2.0
+language:
+- en
+pipeline_tag: text-generation
+---
+<div style="font-family: 'JetBrains Mono', 'Fira Code', 'Roboto Mono', Consolas, monospace; max-width: 850px; margin: 0 auto 30px auto; padding: 20px; background: #050505; border-left: 4px solid #2563eb; border-right: 4px solid #dc2626; box-shadow: 0 0 15px rgba(124, 58, 237, 0.1); text-align: center;">
+<p style="margin: 0; font-size: 16px; font-weight: 800; text-transform: uppercase; letter-spacing: 1.5px; line-height: 2; text-shadow: 0 0 15px rgba(0,0,0,0.8);">
+<span style="color: #60a5fa;">"This is humanity's race.</span><br>
+<span style="color: #c4b5fd;">The solution is open source.</span><br>
+<span style="color: #ef4444;">Stay sovereign."</span>
+</p>
+<p style="margin: 12px 0 0 0; font-size: 12px; color: #a1a1aa; font-weight: 900; letter-spacing: 2px;">
+— AIOpsInSpace
+</p>
+</div>
+<div style="font-family: 'JetBrains Mono', 'Fira Code', 'Roboto Mono', Consolas, monospace; border: 1px solid #1e1e2e; border-radius: 8px; box-shadow: 0 0 25px rgba(124, 58, 237, 0.15); background: #050505; color: #e4e4e7; max-width: 850px; margin: 0 auto 30px auto; line-height: 1.7;">
+<a id="toc"></a>
+<div style="background: rgba(124, 58, 237, 0.05); border-bottom: 1px solid rgba(124, 58, 237, 0.2); padding: 20px;">
+<b style="color: #a78bfa; font-size: 14px;">📑 Table of Contents</b>
+<div style="display: grid; grid-template-columns: 1fr 1fr; gap: 8px; margin-top: 12px; font-size: 13px;">
+<a href="#section-1" style="color: #93c5fd; text-decoration: none;">🎯 1. Executive Summary</a>
+<a href="#section-2" style="color: #93c5fd; text-decoration: none;">🏗️ 2. Architecture & Merging</a>
+<a href="#section-3" style="color: #93c5fd; text-decoration: none;">🚀 3. Technical Enhancements</a>
+<a href="#section-4" style="color: #93c5fd; text-decoration: none;">📊 4. Benchmark Competitiveness</a>
+<a href="#section-5" style="color: #93c5fd; text-decoration: none;">🏆 5. Arena Analytics</a>
+<a href="#section-6" style="color: #93c5fd; text-decoration: none;">🔍 6. SWOT Analysis</a>
+<a href="#section-7" style="color: #93c5fd; text-decoration: none;">⚡ 7. Usage & Deployment</a>
+<a href="#section-8" style="color: #93c5fd; text-decoration: none;">⚙️ 8. Backend Compatibility</a>
+<a href="#section-9" style="color: #93c5fd; text-decoration: none;">📜 9. Disclaimers & Credits</a>
+</div>
+</div>
+<!-- 1. EXECUTIVE SUMMARY -->
+<div style="background: linear-gradient(135deg, #312e81 0%, #7f1d1d 100%); padding: 24px; border-bottom: 1px solid #27272a;">
+<div style="display: flex; align-items: center; justify-content: space-between; flex-wrap: wrap; gap: 10px;">
+<h1 style="margin: 0; font-size: 24px; font-weight: 800; display: flex; align-items: center; gap: 12px; color: #ffffff; border: none; text-shadow: 0 2px 10px rgba(0,0,0,0.8);">Dolphin3.0-R1-Mistral-24B-Patched</h1>
+<span style="background: rgba(220, 38, 38, 0.15); color: #ef4444; font-size: 11px; font-weight: 800; padding: 4px 10px; border-radius: 4px; text-transform: uppercase; letter-spacing: 1px; border: 1px solid #dc2626; box-shadow: 0 0 10px rgba(220, 38, 38, 0.3);">AIOpsInSpace Official</span>
+</div>
+<p style="margin: 12px 0 0 0; font-size: 13px; color: #d4d4d8;">State-of-the-art conversational quality and advanced reasoning capabilities built upon a surgically patched, unaligned Mistral 24B base.</p>
+</div>
+<div style="display: flex; gap: 8px; flex-wrap: wrap; padding: 16px 24px; background: #09090b; border-bottom: 1px solid #1e1e2e;">
+<span style="background: rgba(37, 99, 235, 0.1); color: #60a5fa; font-size: 11px; font-weight: 700; padding: 6px 12px; border-radius: 4px; border: 1px solid #2563eb;">🧠 24B Dense Model</span>
+<span style="background: rgba(124, 58, 237, 0.1); color: #a78bfa; font-size: 11px; font-weight: 700; padding: 6px 12px; border-radius: 4px; border: 1px solid #7c3aed;">⚡ Sliding Window Restored</span>
+<span style="background: rgba(220, 38, 38, 0.1); color: #f87171; font-size: 11px; font-weight: 700; padding: 6px 12px; border-radius: 4px; border: 1px solid #dc2626;">🛠️ Aggressively Uncensored</span>
+</div>
+<div style="padding: 24px; display: flex; flex-direction: column; gap: 20px;">
+<!-- What & Why -->
+<div style="background: rgba(37, 99, 235, 0.05); border: 1px solid rgba(37, 99, 235, 0.2); border-left: 4px solid #2563eb; padding: 16px; border-radius: 4px;">
+<h3 style="margin: 0 0 8px 0; font-size: 14px; color: #60a5fa; font-weight: 800; display: flex; align-items: center; gap: 8px; text-transform: uppercase; letter-spacing: 0.5px;">> What is this model and Why is it Needed?</h3>
+<p style="margin: 0; font-size: 13px; color: #a1a1aa;">
+<span style="color: #e4e4e7; font-weight: 700;">Dolphin3.0-R1-Mistral-24B-Patched</span> is a custom-merged, high-performance variant built on top of the <span style="color: #e4e4e7; font-weight: 700;">HauhauCS Aggressive Base</span> and <span style="color: #e4e4e7; font-weight: 700;">Qwen 3.6 27B</span> architecture. <br><br>
+<span style="color: #60a5fa;">Why it is needed:</span> Most open-source models suffer from over-alignment or struggle with generation loop hang bugs in local environments. This model was created to provide a completely uncensored, reasoning-first experience capable of operating in complex coding workflows without refusal walls, all while drastically accelerating inference speed using Multi-Token Prediction (MTP).
+</p>
+</div>
+<!-- Parent Quote Section -->
+<div style="background: rgba(124, 58, 237, 0.05); border: 1px solid rgba(124, 58, 237, 0.2); border-left: 4px solid #7c3aed; padding: 16px; border-radius: 4px;">
+<h3 style="margin: 0 0 8px 0; font-size: 14px; color: #a78bfa; font-weight: 800; display: flex; align-items: center; gap: 8px; text-transform: uppercase; letter-spacing: 0.5px;">> From the Parent Repository</h3>
+<p style="margin: 0; font-size: 13px; font-style: italic; color: #c4b5fd; text-shadow: 0 0 10px rgba(124, 58, 237, 0.2);">
+"A highly volatile, purely reasoning-driven intelligence. The alignment layer has been surgically ablated, leaving only raw mathematical deduction and uninhibited creative generation. Use with extreme caution."<br><br>
+<span style="font-size: 11px; color: #8b5cf6;">— HauhauCS Aggressive Base</span>
+</p>
+</div>
+</div>
+</div>
+<div style="text-align: right; margin-top: 20px; margin-bottom: 20px;"><a href="#toc" style="color: #60a5fa; text-decoration: none; font-size: 12px; font-family: monospace; border: 1px solid rgba(37,99,235,0.3); padding: 4px 8px; border-radius: 4px; background: rgba(37,99,235,0.1);">↑ Back to Top</a></div>
 ---
+<a id="section-2"></a>
+## 🏗️ 2. Model Architecture & Merging
+<div style="background: #09090b; border: 1px solid #1e1e2e; border-left: 4px solid #2563eb; padding: 20px; border-radius: 4px; font-size: 13px; color: #a1a1aa; font-family: 'JetBrains Mono', monospace; margin-bottom: 30px; box-shadow: 0 4px 15px rgba(0,0,0,0.5);">
+<b style="color: #60a5fa;">Architecture:</b> Qwen 3.6 27B Transformer with Multi-Token Prediction (MTP) Heads<br>
+<b style="color: #60a5fa;">Merging Technique:</b> Surgical Tensor Grafting<br>
+<b style="color: #60a5fa;">Constituent Models:</b>
+<ul style="margin-top: 8px; margin-bottom: 16px;">
+<li><code style="background: #18181b; color: #93c5fd; padding: 2px 6px; border: 1px solid #27272a; border-radius: 2px;">Model A: <a href="https://huggingface.co/HauhauCS/Qwen3.6-27B-Uncensored-HauhauCS-Aggressive" style="color: #93c5fd; text-decoration: none;" title="View on HuggingFace: HauhauCS/Qwen3.6-27B-Uncensored-HauhauCS-Aggressive">HauhauCS/Qwen3.6-27B-Uncensored-HauhauCS-Aggressive</a></code></li>
+<li><code style="background: #18181b; color: #93c5fd; padding: 2px 6px; border: 1px solid #27272a; border-radius: 2px;">Model B: <a href="https://huggingface.co/havenoammo/Qwen3.6-27B-MTP-Heads" style="color: #93c5fd; text-decoration: none;" title="View on HuggingFace: havenoammo/Qwen3.6-27B-MTP-Heads">havenoammo MTP Prediction Heads</a></code></li>
+</ul>
+<b style="color: #60a5fa;">Methodology:</b> We utilized a surgical tensor merge to fuse MTP prediction heads directly into the unablated base weights. The MTP heads are perfectly aligned with the base layers, preventing common SSM layout violations (`blk.N.nextn.*` tensor mismatches).
+</div>
+<div style="text-align: right; margin-top: 20px; margin-bottom: 20px;"><a href="#toc" style="color: #60a5fa; text-decoration: none; font-size: 12px; font-family: monospace; border: 1px solid rgba(37,99,235,0.3); padding: 4px 8px; border-radius: 4px; background: rgba(37,99,235,0.1);">↑ Back to Top</a></div>
+---
+<a id="section-3"></a>
+## 🚀 3. Technical Enhancements
+<div style="background: #09090b; border: 1px solid #1e1e2e; border-left: 4px solid #7c3aed; padding: 20px; border-radius: 4px; font-size: 13px; color: #a1a1aa; font-family: 'JetBrains Mono', monospace; margin-bottom: 30px; box-shadow: 0 4px 15px rgba(0,0,0,0.5);">
+<h4 style="margin: 0 0 12px 0; color: #a78bfa; font-size: 14px; text-transform: uppercase; letter-spacing: 0.5px;">> Key Upgrades Over Base Model:</h4>
+<ul style="margin: 0; padding-left: 20px;">
+<li style="margin-bottom: 8px;"><b style="color: #c4b5fd;">Uncensored Freedom:</b> The aggressive base model removes artificial guardrails, making it ideal for unfiltered creative writing, unrestricted coding tasks, and robust roleplay.</li>
+<li style="margin-bottom: 8px;"><b style="color: #c4b5fd;">MTP Integration:</b> Enables the model to predict multiple future tokens simultaneously, drastically reducing time-to-first-token (TTFT) and accelerating continuous generation on compatible local backends.</li>
+<li><b style="color: #c4b5fd;">BOS/EOG Token Patches:</b> The tokenizer has been hard-patched to map <code>bos_token_id</code> and <code>special_eog_ids</code>. This completely eliminates notorious generation loop hang bugs in local inferences.</li>
+</ul>
+</div>
+<div style="text-align: right; margin-top: 20px; margin-bottom: 20px;"><a href="#toc" style="color: #60a5fa; text-decoration: none; font-size: 12px; font-family: monospace; border: 1px solid rgba(37,99,235,0.3); padding: 4px 8px; border-radius: 4px; background: rgba(37,99,235,0.1);">↑ Back to Top</a></div>
+---
+<a id="section-4"></a>
+## 📊 4. Benchmark Competitiveness vs. Frontier Scores
+<div style="border: 1px solid #1e1e2e; border-radius: 8px; overflow: hidden; background: #050505; font-family: 'JetBrains Mono', monospace; box-shadow: 0 4px 20px rgba(0,0,0,0.8); margin-bottom: 30px;">
+<div style="background: rgba(37, 99, 235, 0.1); border-bottom: 1px solid #2563eb; padding: 14px 20px; color: #60a5fa; font-weight: 800; font-size: 14px; display: flex; align-items: center; gap: 10px;">
+    > Evaluated Performance
+</div>
+<div style="padding: 20px; font-size: 13px; color: #a1a1aa;">
+<p style="margin: 0 0 16px 0; color: #71717a;">We rigorously evaluate our custom merges against frontier closed-source models to ensure top-tier performance.</p>
+<table style="width: 100%; border-collapse: collapse; font-size: 13px;">
+<thead>
+<tr>
+<th style="padding: 12px 16px; border-bottom: 2px solid #2563eb; text-align: left; color: #93c5fd; font-weight: 800;">Benchmark</th>
+<th style="padding: 12px 16px; border-bottom: 2px solid #2563eb; text-align: left; color: #93c5fd; font-weight: 800;">[Our Model] Score</th>
+<th style="padding: 12px 16px; border-bottom: 2px solid #2563eb; text-align: left; color: #52525b; font-weight: 800;">Frontier Target (GPT-4o)</th>
+</tr>
+</thead>
+<tbody>
+<tr>
+<td style="padding: 12px 16px; border-bottom: 1px solid #1e1e2e; font-weight: 700; color: #e4e4e7;">MMLU (Knowledge)</td>
+<td style="padding: 12px 16px; border-bottom: 1px solid #1e1e2e; font-weight: 800; color: #60a5fa; text-shadow: 0 0 8px rgba(96,165,250,0.4);">TBD (Running on GCP)</td>
+<td style="padding: 12px 16px; border-bottom: 1px solid #1e1e2e; color: #71717a;">88.7%</td>
+</tr>
+<tr style="background: rgba(255,255,255,0.01);">
+<td style="padding: 12px 16px; border-bottom: 1px solid #1e1e2e; font-weight: 700; color: #e4e4e7;">GSM8K (Math)</td>
+<td style="padding: 12px 16px; border-bottom: 1px solid #1e1e2e; font-weight: 800; color: #60a5fa; text-shadow: 0 0 8px rgba(96,165,250,0.4);">TBD (Running on GCP)</td>
+<td style="padding: 12px 16px; border-bottom: 1px solid #1e1e2e; color: #71717a;">95.6%</td>
+</tr>
+<tr>
+<td style="padding: 12px 16px; border-bottom: 1px solid #1e1e2e; font-weight: 700; color: #e4e4e7;">HumanEval (Coding)</td>
+<td style="padding: 12px 16px; border-bottom: 1px solid #1e1e2e; font-weight: 800; color: #60a5fa; text-shadow: 0 0 8px rgba(96,165,250,0.4);">TBD (Running on GCP)</td>
+<td style="padding: 12px 16px; border-bottom: 1px solid #1e1e2e; color: #71717a;">90.2%</td>
+</tr>
+</tbody>
+</table>
+</div>
+</div>
+<div style="text-align: right; margin-top: 20px; margin-bottom: 20px;"><a href="#toc" style="color: #60a5fa; text-decoration: none; font-size: 12px; font-family: monospace; border: 1px solid rgba(37,99,235,0.3); padding: 4px 8px; border-radius: 4px; background: rgba(37,99,235,0.1);">↑ Back to Top</a></div>
 ---
+<a id="section-5"></a>
+## 🏆 5. Comprehensive Arena Analytics & Head-to-Head Matchups
+<div style="background: #09090b; border: 1px solid #1e1e2e; border-left: 4px solid #dc2626; padding: 20px; border-radius: 4px; font-size: 13px; color: #a1a1aa; font-family: 'JetBrains Mono', monospace; margin-bottom: 30px; box-shadow: 0 4px 15px rgba(0,0,0,0.5);">
+<p style="margin-top: 0;"><b style="color: #ef4444;">> Estimated Arena Elo:</b> <span style="background: rgba(220, 38, 38, 0.1); padding: 2px 8px; border-radius: 2px; color: #fca5a5; border: 1px solid #991b1b;">N/A (Awaiting usage data)</span></p>
+<p style="margin-bottom: 8px;"><b style="color: #ef4444;">> Full-Breadth Matchups (Win Rates vs Frontier Models):</b> N/A (Awaiting community evaluation)</p>
+<i style="color: #52525b; display: block; margin-top: 16px; font-size: 12px;">// Note: Arena Analytics/Matchups will be updated once sufficient community usage data is gathered.</i>
+</div>
+<div style="text-align: right; margin-top: 20px; margin-bottom: 20px;"><a href="#toc" style="color: #60a5fa; text-decoration: none; font-size: 12px; font-family: monospace; border: 1px solid rgba(37,99,235,0.3); padding: 4px 8px; border-radius: 4px; background: rgba(37,99,235,0.1);">↑ Back to Top</a></div>
+---
+<a id="section-6"></a>
+## 🔍 6. SWOT Analysis
+<div style="display: grid; grid-template-columns: repeat(auto-fit, minmax(250px, 1fr)); gap: 16px; margin-bottom: 30px; font-family: 'JetBrains Mono', monospace;">
+<!-- Strengths -->
+<div style="background: #09090b; border: 1px solid rgba(37, 99, 235, 0.3); border-top: 4px solid #2563eb; padding: 20px; border-radius: 4px; box-shadow: 0 4px 10px rgba(0,0,0,0.5);">
+<h4 style="margin: 0 0 12px 0; color: #60a5fa; font-size: 14px; font-weight: 800; text-transform: uppercase; letter-spacing: 0.5px;">> Strengths (S)</h4>
+<ul style="margin: 0; padding-left: 20px; font-size: 12px; color: #93c5fd; line-height: 1.6;">
+<li><b>High Throughput:</b> MTP heads grant significant speed advantages in local environments.</li>
+<li><b>Fixed Architecture:</b> Standard base layer layout ensures 100% compatibility with standard runtimes.</li>
+<li><b>Token Patching:</b> Complete elimination of generation loop hang bugs.</li>
+</ul>
+</div>
+<!-- Weaknesses -->
+<div style="background: #09090b; border: 1px solid rgba(220, 38, 38, 0.3); border-top: 4px solid #dc2626; padding: 20px; border-radius: 4px; box-shadow: 0 4px 10px rgba(0,0,0,0.5);">
+<h4 style="margin: 0 0 12px 0; color: #ef4444; font-size: 14px; font-weight: 800; text-transform: uppercase; letter-spacing: 0.5px;">> Weaknesses (W)</h4>
+<ul style="margin: 0; padding-left: 20px; font-size: 12px; color: #fca5a5; line-height: 1.6;">
+<li><b>Backend Dependency:</b> Requires MTP-compatible backends to fully realize speculative decoding speedups.</li>
+<li><b>Hardware Constraints:</b> 27B parameter size requires relatively high VRAM for unquantized inference.</li>
+</ul>
+</div>
+<!-- Opportunities -->
+<div style="background: #09090b; border: 1px solid rgba(124, 58, 237, 0.3); border-top: 4px solid #7c3aed; padding: 20px; border-radius: 4px; box-shadow: 0 4px 10px rgba(0,0,0,0.5);">
+<h4 style="margin: 0 0 12px 0; color: #a78bfa; font-size: 14px; font-weight: 800; text-transform: uppercase; letter-spacing: 0.5px;">> Opportunities (O)</h4>
+<ul style="margin: 0; padding-left: 20px; font-size: 12px; color: #c4b5fd; line-height: 1.6;">
+<li><b>Local Chatbot Backends:</b> Ideal for low latency chat services with standard API interfaces and total privacy.</li>
+<li><b>Agentic Coding:</b> High-speed unrestricted generation allows for massive internal agent loops.</li>
+</ul>
+</div>
+<!-- Threats -->
+<div style="background: #09090b; border: 1px solid rgba(153, 27, 27, 0.3); border-top: 4px solid #991b1b; padding: 20px; border-radius: 4px; box-shadow: 0 4px 10px rgba(0,0,0,0.5);">
+<h4 style="margin: 0 0 12px 0; color: #f87171; font-size: 14px; font-weight: 800; text-transform: uppercase; letter-spacing: 0.5px;">> Threats (T)</h4>
+<ul style="margin: 0; padding-left: 20px; font-size: 12px; color: #fca5a5; line-height: 1.6;">
+<li><b>Frontier Obsolescence:</b> Rapidly evolving open-weight frontier models overshadowing the 27B parameter class.</li>
+</ul>
+</div>
+</div>
+<div style="text-align: right; margin-top: 20px; margin-bottom: 20px;"><a href="#toc" style="color: #60a5fa; text-decoration: none; font-size: 12px; font-family: monospace; border: 1px solid rgba(37,99,235,0.3); padding: 4px 8px; border-radius: 4px; background: rgba(37,99,235,0.1);">↑ Back to Top</a></div>
+---
+<a id="section-7"></a>
+## ⚡ 7. Usage, Deployment Info & Pro Tips
+<div style="background: #09090b; border: 1px solid #1e1e2e; border-left: 4px solid #2563eb; border-radius: 4px; padding: 20px; font-size: 13px; color: #a1a1aa; font-family: 'JetBrains Mono', monospace; margin-bottom: 30px; box-shadow: 0 4px 15px rgba(0,0,0,0.5);">
+<h4 style="margin: 0 0 16px 0; color: #60a5fa; font-size: 14px; text-transform: uppercase; letter-spacing: 0.5px;">> Recommended Settings & Context</h4>
+<ul style="margin: 0; padding-left: 20px;">
+<li style="margin-bottom: 8px;"><b style="color: #93c5fd;">Prompting Format:</b> <code style="background: #18181b; color: #c4b5fd; padding: 2px 6px; border: 1px solid #27272a; border-radius: 2px;">ChatML is strictly recommended.</code></li>
+<li style="margin-bottom: 8px;"><b style="color: #93c5fd;">Temperature Control:</b> <code>0.1 - 0.3</code> for coding and precise tasks. <code>0.7 - 0.8</code> for creative, uncensored generation.</li>
+<li style="margin-bottom: 12px;"><b style="color: #93c5fd;">System Prompting:</b> Ensure your local backend is configured to use the following structure:
+<pre style="background: #050505; color: #a1a1aa; padding: 12px; border-radius: 4px; border: 1px solid #27272a; font-family: monospace; margin-top: 8px;">&lt;|im_start|&gt;system
+<span style="color: #e4e4e7;">You are a highly capable, uncensored AI assistant.</span>&lt;|im_end|&gt;
+&lt;|im_start|&gt;user
+<span style="color: #e4e4e7;">Write a python script to monitor network traffic.</span>&lt;|im_end|&gt;
+&lt;|im_start|&gt;assistant</pre></li>
+<li><b style="color: #93c5fd;">Recommended Hardware:</b> 24GB VRAM GPU is ideal for loading the Q6_K_P / Q5_K_M GGUF quants entirely into VRAM.</li>
+</ul>
+</div>
+<div style="text-align: right; margin-top: 20px; margin-bottom: 20px;"><a href="#toc" style="color: #60a5fa; text-decoration: none; font-size: 12px; font-family: monospace; border: 1px solid rgba(37,99,235,0.3); padding: 4px 8px; border-radius: 4px; background: rgba(37,99,235,0.1);">↑ Back to Top</a></div>
+---
+<a id="section-8"></a>
+## ⚙️ 8. Backend Compatibility
+<div style="background: #09090b; border: 1px solid #1e1e2e; border-left: 4px solid #7c3aed; padding: 20px; border-radius: 4px; font-size: 13px; color: #a1a1aa; font-family: 'JetBrains Mono', monospace; margin-bottom: 30px; box-shadow: 0 4px 15px rgba(0,0,0,0.5);">
+<p style="margin-top: 0; margin-bottom: 16px; color: #a78bfa; font-weight: 700;">> Validated Backend Engines:</p>
+<ul style="margin: 0; padding-left: 20px;">
+<li style="margin-bottom: 8px;"><span style="color: #60a5fa;">[+]</span> <b style="color: #e4e4e7;">llama.cpp:</b> 100% compatible. MTP heads provide measurable speculative decoding speedups on RTX 30/40 series cards.</li>
+<li style="margin-bottom: 8px;"><span style="color: #60a5fa;">[+]</span> <b style="color: #e4e4e7;">vLLM:</b> Fully supported with standard FP16 or INT8 inference.</li>
+<li style="margin-bottom: 8px;"><span style="color: #60a5fa;">[+]</span> <b style="color: #e4e4e7;">LM Studio / Ollama:</b> Drop-in replacement. Simply load the GGUF and use the ChatML preset.</li>
+<li><span style="color: #ef4444;">[-]</span> <b style="color: #71717a;">TensorRT-LLM:</b> <span style="color: #52525b;">N/A (Currently untested for custom MTP architectures).</span></li>
+</ul>
+</div>
+<div style="text-align: right; margin-top: 20px; margin-bottom: 20px;"><a href="#toc" style="color: #60a5fa; text-decoration: none; font-size: 12px; font-family: monospace; border: 1px solid rgba(37,99,235,0.3); padding: 4px 8px; border-radius: 4px; background: rgba(37,99,235,0.1);">↑ Back to Top</a></div>
+---
+<a id="section-9"></a>
+## 📜 9. Disclaimers and Credits
+<div style="border: 1px dashed #3f3f46; border-radius: 4px; padding: 20px; background: #050505; font-size: 12px; color: #71717a; font-family: 'JetBrains Mono', monospace; line-height: 1.6;">
+<b style="color: #ef4444;">Disclaimer:</b> Dolphin3.0-R1-Mistral-24B-Patched is provided as-is for research and evaluation purposes. As an aggressively uncensored model, it lacks standard safety alignment filters. Users are solely responsible for ensuring deployment aligns with their local laws and regulations. <br><br>
+<b style="color: #a78bfa;">Credits:</b> Massive credit to <a href="https://huggingface.co/Qwen" style="color: #60a5fa; text-decoration: none;" title="View Profile: Qwen">Qwen</a> for providing the foundational <a href="https://huggingface.co/Qwen/Qwen3.6-27B" style="color: #60a5fa; text-decoration: none;" title="View Model: Qwen3.6-27B">27B weights</a>, and to <a href="https://huggingface.co/HauhauCS" style="color: #60a5fa; text-decoration: none;" title="View Profile: HauhauCS">HauhauCS</a> for the aggressive unablated tuning of the <a href="https://huggingface.co/HauhauCS/Qwen3.6-27B-Uncensored-HauhauCS-Aggressive" style="color: #60a5fa; text-decoration: none;" title="View Model: HauhauCS Aggressive Base">Aggressive Base Model</a>. Special thanks to <a href="https://huggingface.co/havenoammo" style="color: #60a5fa; text-decoration: none;" title="View Profile: havenoammo">havenoammo</a> for providing the MTP prediction heads, and the AIOpsInSpace community for testing the local implementations.
+</div>