Instructions to use open-chess/Caissa-Chess-M1-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use open-chess/Caissa-Chess-M1-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="open-chess/Caissa-Chess-M1-GGUF",
	filename="Caissa-Chess-M1-F16.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 open-chess/Caissa-Chess-M1-GGUF with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf open-chess/Caissa-Chess-M1-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf open-chess/Caissa-Chess-M1-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf open-chess/Caissa-Chess-M1-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf open-chess/Caissa-Chess-M1-GGUF:Q4_K_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 open-chess/Caissa-Chess-M1-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf open-chess/Caissa-Chess-M1-GGUF:Q4_K_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 open-chess/Caissa-Chess-M1-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf open-chess/Caissa-Chess-M1-GGUF:Q4_K_M

Use Docker

docker model run hf.co/open-chess/Caissa-Chess-M1-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use open-chess/Caissa-Chess-M1-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "open-chess/Caissa-Chess-M1-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "open-chess/Caissa-Chess-M1-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/open-chess/Caissa-Chess-M1-GGUF:Q4_K_M

Ollama
How to use open-chess/Caissa-Chess-M1-GGUF with Ollama:
```
ollama run hf.co/open-chess/Caissa-Chess-M1-GGUF:Q4_K_M
```

Unsloth Studio

How to use open-chess/Caissa-Chess-M1-GGUF 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 open-chess/Caissa-Chess-M1-GGUF 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 open-chess/Caissa-Chess-M1-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for open-chess/Caissa-Chess-M1-GGUF to start chatting

How to use open-chess/Caissa-Chess-M1-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf open-chess/Caissa-Chess-M1-GGUF:Q4_K_M

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": "open-chess/Caissa-Chess-M1-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use open-chess/Caissa-Chess-M1-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf open-chess/Caissa-Chess-M1-GGUF:Q4_K_M

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 open-chess/Caissa-Chess-M1-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat new
Docker Model Runner
How to use open-chess/Caissa-Chess-M1-GGUF with Docker Model Runner:
```
docker model run hf.co/open-chess/Caissa-Chess-M1-GGUF:Q4_K_M
```

Lemonade

How to use open-chess/Caissa-Chess-M1-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull open-chess/Caissa-Chess-M1-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Caissa-Chess-M1-GGUF-Q4_K_M

List all available models

lemonade list

Caissa-Chess-M1-GGUF

📄 Overview


Base Model	open-chess/Caissa-Chess-M1
Parameters	8B
Dataset	MetaChess-20k

Quant types

Quant type	Size	Recommended Use
Q2_K	3.02 GB	Ultra-low RAM (mobile, embedded)
Q3_K_S	3.49 GB	Low RAM devices
Q3_K_M	3.81 GB	Low RAM devices
Q4_K_S	4.46 GB	Balanced quality/size
Q4_K_M	4.68 GB	Recommended — best balance
Q5_K_S	5.32 GB	Higher quality
Q5_K_M	5.44 GB	Higher quality
Q6_K	6.25 GB	Near-original quality
Q8_0	8.1 GB	Maximum quality
F16	15.2 GB	Original precision

🎯 Intended Use

This model is designed for chess position analysis with structured Chain-of-Thought reasoning. It is optimized for:

Chess analysis — evaluating positions, calculating variations, finding best moves
Educational applications — explaining chess concepts and strategic thinking
On-device chess assistants — runs on mobile, Raspberry Pi, or CPU-only environments
Chess AI research — studying reasoning patterns in small language models
Local inference — privacy-focused chess analysis without cloud APIs

Not recommended for: general-purpose reasoning, non-chess tasks, or production systems requiring 100% move accuracy.

💬 Example Usage

Using llama.cpp

./llama-cli -m Caissa-Chess-M1-Q4_K_M.gguf \
  -p "You are a chess grandmaster and analyst. Your task is to deeply analyze a position and find the best move. You MUST reason in <think> tags, then output the move in <move> tag.

FEN: r1bqkbnr/pppp1ppp/2n5/4p3/4P3/5N2/PPPP1PPP/RNBQKB1R w KQkq - 2 3" \
  -n 512 -t 4

Expected Output Format:

<think>
<evaluation>Position is roughly equal (+0.00 pawns). It's White's turn to move.</evaluation>
<calculation>Depth: 60. Principal variation: d2d4 -> c8f5 -> f1e2 -> g8f6 -> e1h1 -> f8e7</calculation>
<reasoning>The move d2d4 establishes a classical center, claiming space and preparing to challenge Black's control of the e5 square...</reasoning>
</think>
<move>d2d4</move>

⚠️ Limitations

Chess-specific — trained exclusively on chess positions; general reasoning or non-chess tasks will be suboptimal
Size constraints — 8B parameters, so extremely complex positions (20+ move calculations) may be simplified
No multimodal — text-only FEN input; no image/board recognition
Move accuracy — may occasionally suggest suboptimal moves; always verify with Stockfish for critical analysis
Training data — trained on 20,000 positions from Lichess (depth 28+); may not handle rare openings well

📖 Citation

@misc{Caissa-Chess-M1-GGUF, author = {Open Chess AI}, title = {Caissa-Chess-M1-GGUF: Chess Reasoning Model with Chain-of-Thought}, year = {2026}, publisher = {Hugging Face}, howpublished = {\url{https://huggingface.co/open-chess/Caissa-Chess-M1-GGUF}}, }

Made by OpenChess, an open source chess AI research project ❤️

Downloads last month: 511

GGUF

Model size

8B params

Architecture

qwen2

Hardware compatibility

2-bit

3-bit

4-bit

5-bit

6-bit

8-bit

16-bit

Model tree for open-chess/Caissa-Chess-M1-GGUF

Base model

Qwen/Qwen2.5-7B

Finetuned

Qwen/Qwen2.5-7B-Instruct

Finetuned

open-chess/Caissa-Chess-M1

Quantized

(2)

this model

open-chess
/

Caissa-Chess-M1-GGUF