jc-builds/SmolLM2-135M-Instruct-Q4_K_M-GGUF

SmolLM2-135M-Instruct Q4_K_M GGUF

This is a Q4_K_M quantized GGUF conversion of HuggingFaceTB/SmolLM2-135M-Instruct optimized for on-device inference with llama.cpp.

Model Details

Property	Value
Original Model	SmolLM2-135M-Instruct
Parameters	135 million
Quantization	Q4_K_M (4-bit, medium quality)
File Size	~101 MB
Context Window	8,192 tokens
Architecture	LLaMA-style transformer
Training Data	2 trillion tokens

Intended Use

This model is optimized for:

• Mobile/Edge Deployment: Runs on ANY iOS device, even the oldest
• llama.cpp Integration: Compatible with llama.cpp and its bindings
• On-Device AI: Private, offline inference without cloud dependencies

Capabilities

• Ultra-Tiny: Smallest model available, instant responses
• Works Everywhere: Runs on any device with minimal resources
• Basic Q&A: Good for simple chat and quick interactions
• Minimal Battery Usage: Extremely efficient
• Trained on 2T Tokens: Impressive capability for its tiny size

Usage with llama.cpp

./llama-cli -m SmolLM2-135M-Instruct.Q4_K_M.gguf -p "Your prompt here" -n 512

License

This model inherits the Apache 2.0 license from the original SmolLM2 model.

Attribution

• Original Model: SmolLM2-135M-Instruct by Hugging Face
• Quantization: jc-builds