Instructions to use badtheorylabs/BTL-3-Compact with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use badtheorylabs/BTL-3-Compact 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 badtheorylabs/BTL-3-Compact
# Run inference directly in the terminal:
llama cli -hf badtheorylabs/BTL-3-Compact

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf badtheorylabs/BTL-3-Compact
# Run inference directly in the terminal:
llama cli -hf badtheorylabs/BTL-3-Compact

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 badtheorylabs/BTL-3-Compact
# Run inference directly in the terminal:
./llama-cli -hf badtheorylabs/BTL-3-Compact

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 badtheorylabs/BTL-3-Compact
# Run inference directly in the terminal:
./build/bin/llama-cli -hf badtheorylabs/BTL-3-Compact

Use Docker

docker model run hf.co/badtheorylabs/BTL-3-Compact

LM Studio
Jan

vLLM

How to use badtheorylabs/BTL-3-Compact with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "badtheorylabs/BTL-3-Compact"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "badtheorylabs/BTL-3-Compact",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/badtheorylabs/BTL-3-Compact

Ollama
How to use badtheorylabs/BTL-3-Compact with Ollama:
```
ollama run hf.co/badtheorylabs/BTL-3-Compact
```

Unsloth Studio

How to use badtheorylabs/BTL-3-Compact 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 badtheorylabs/BTL-3-Compact 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 badtheorylabs/BTL-3-Compact to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for badtheorylabs/BTL-3-Compact to start chatting

How to use badtheorylabs/BTL-3-Compact with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf badtheorylabs/BTL-3-Compact

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "badtheorylabs/BTL-3-Compact"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use badtheorylabs/BTL-3-Compact with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf badtheorylabs/BTL-3-Compact

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default badtheorylabs/BTL-3-Compact

Run Hermes

hermes

Atomic Chat new

OpenClaw new

How to use badtheorylabs/BTL-3-Compact with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf badtheorylabs/BTL-3-Compact

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "badtheorylabs/BTL-3-Compact" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Docker Model Runner
How to use badtheorylabs/BTL-3-Compact with Docker Model Runner:
```
docker model run hf.co/badtheorylabs/BTL-3-Compact
```

Lemonade

How to use badtheorylabs/BTL-3-Compact with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull badtheorylabs/BTL-3-Compact

Run and chat with the model

lemonade run user.BTL-3-Compact-{{QUANT_TAG}}

List all available models

lemonade list

BTL-3-Compact

File size: 2,459 Bytes

{
  "schema_version": 1,
  "release": "BTL-3 Compact — Thinking Escape r8 step160 v3 canonical refresh",
  "checkpoint": "RL-0013",
  "architecture": "Qwen3.6-27B",
  "model": {
    "path": "model/BTL-3-Compact-AVQ2.gguf",
    "bytes": 8392369600,
    "sha256": "0a4d9ddee49e5aa93586a792bd4d452ea837229d49d22e54212dde87a5c9888a"
  },
  "runtimes": {
    "supported": [
      {
        "name": "BTL-3-Compact-macos-arm64",
        "path": "runtimes/supported/BTL-3-Compact-macos-arm64",
        "bundle": "BTL-3 Compact macOS arm64",
        "status": "verified native runtime"
      }
    ],
    "preview": [
      {
        "name": "BTL-3-Compact-linux-arm64-cuda",
        "path": "runtimes/preview/BTL-3-Compact-linux-arm64-cuda",
        "bundle": "BTL-3 Compact Linux arm64 CUDA (DGX Spark)",
        "status": "cross-compiled; NVIDIA runtime conformance pending"
      }
    ]
  },
  "integrations": [
    "integrations/btl3-native",
    "integrations/ollama"
  ],
  "documentation": [
    "docs/launch-btl3-compact.md",
    "docs/launch-btl3-cuda.md",
    "docs/patched-ollama-and-lmstudio.md",
    "docs/btl-3-compact-gguf-exporter.md"
  ],
  "evidence": [
    "evidence/BTL-3-Compact-AVQ2.report.json",
    "evidence/native-load-report.json",
    "evidence/rtx-pro-6000-speed-2026-07-19.md",
    "evidence/compact-validation.md",
    "evidence/thinking-escape-v3-smoke.json",
    "evidence/thinking-escape-v3-tool-gate.json",
    "evidence/thinking-escape-v3-native-runtime.json",
    "evidence/thinking-escape-v3-training-manifest.json"
  ],
  "tools": [
    "tools/install_consumer_bundle.py"
  ],
  "licenses": [
    "licenses/LICENSE",
    "licenses/LICENSE.runtime",
    "licenses/THIRD_PARTY_NOTICES.md"
  ],
  "stock_ollama_compatible": false,
  "stock_lm_studio_engine_compatible": false,
  "default_thinking": false,
  "canonical_refresh": {
    "date": "2026-07-24",
    "behavior_adapter": {
      "identity": "thinking-escape-r8-step160-v3",
      "rank": 8,
      "alpha": 16,
      "tensor_count": 132,
      "payload_bytes": 32440320,
      "sha256": "7ba9f80be97287775575022e84e735ac3c38b343c0bb83c5c3b0a86c53d6c881"
    },
    "non_behavior_tensors_verified_unchanged": 2284,
    "packed_decoder_unchanged": true,
    "rank32_output_head_correction_unchanged": true,
    "tool_retention": "62/63",
    "thinking_final_answer_rate": "5/12",
    "thinking_coding_termination": "0/3",
    "approved_as_full_thinking_fix": false
  }
}