jc-builds/Ministral-3-3B-Q4_K_M-GGUF

Ministral-3 3B Q4_K_M GGUF

This is a Q4_K_M quantized GGUF conversion of mistralai/Ministral-3B-Instruct-2412 optimized for on-device inference with llama.cpp.

Model Details

Property	Value
Original Model	Ministral-3B-Instruct
Parameters	3 billion
Quantization	Q4_K_M (4-bit, medium quality)
File Size	~2.0 GB
Context Window	256,000 tokens
Architecture	Mistral Transformer

Intended Use

This model is optimized for:

• Mobile/Edge Deployment: Runs efficiently on iOS devices with 6GB+ RAM
• llama.cpp Integration: Compatible with llama.cpp and its bindings
• On-Device AI: Private, offline inference without cloud dependencies

Capabilities

• Fast Inference: Edge-optimized for quick responses
• Massive Context: 256K token context window
• Multilingual: Supports 12+ languages
• Code Generation: Strong coding capabilities
• General Assistant: Versatile for various tasks

Usage with llama.cpp

./llama-cli -m Ministral-3-3B-Q4_K_M.gguf -p "Your prompt here" -n 512

License

This model is released under the Apache 2.0 License.

Attribution

• Original Model: Ministral-3B-Instruct by Mistral AI
• Quantization: jc-builds