Qwen3-1.7B · in-browser WebLLM + WebGPU · 100% local, no server

Idle. Click "Load model" to download Qwen3-1.7B (~1.4 GB, cached after first load).

Ask Qwen3-1.7B a math, coding, or reasoning problem. The model runs entirely in your browser via WebGPU.

Conversation history is kept in this page and replayed to the model each turn.

Dev log