Pick a language model to preload and chat with. Every option runs entirely in your browser — nothing is uploaded anywhere.
Qwen3-0.6B
WebGPU
~340 MB
q4f16
More capable, general-purpose chat. Requires a WebGPU-capable browser — Chrome or Edge on desktop.
Gemma 3 1B
WebGPU
~600 MB
q4f16
Google's compact, newest-generation model — balanced quality and speed. Requires a WebGPU-capable browser — Chrome or Edge on desktop.
Phi-4-mini
WebGPU
~2.2 GB
q4f16
Microsoft's largest option here — the biggest download by far, but the most capable answers. Requires a WebGPU-capable browser — Chrome or Edge on desktop.
SmolLM2-135M-Instruct
WASM CPU
~134 MB
int8
Much smaller and faster to load, runs on any modern browser via WebAssembly — no GPU needed, simpler answers. The same model used by the AI Inference Benchmark.