SmolLM2 135M Instruct

Fastest default and the lightest general chat option.

SmolLM2 360M Instruct

Still small, but noticeably stronger than the 135M model.

SmolLM2 1.7B Instruct

Higher quality general chat, but much heavier to load.

Qwen2.5 0.5B Instruct

Compact general-purpose Qwen instruct model with ONNX weights.

Qwen2.5 Coder 0.5B Instruct

Better fit for code-oriented prompts and structured answers.

Qwen3 0.6B ONNX

Strong browser-ready ONNX model with explicit WebGPU guidance.

Gemma 3 270M IT ONNX

Small Gemma instruction model for lightweight chat experiments.

Granite 4.0 350M ONNX Web

Small multilingual Granite model tuned for browser/WebGPU use.

Bonsai 1.7B ONNX

Large browser model optimized around aggressive WebGPU quantization.

Supra 50M Instruct ONNX

Ultra-small experimental option for the fastest first-load experience.