FrogNano-4B on WebGPU

microsoft/FrogNano-4B-2609, running fully in your browser with Transformers.js (q4f16 ONNX, about 2.4 GB download, cached after the first load). A compact coding model; this is plain chat without its agent tools.

Checking WebGPU...

Ask a coding question once the model is loaded.

Settings

System prompt Enable thinking Max new tokens: 2048 Temperature: 0.6 Top-p: 0.95 Top-k: 20