Web Decision Model Playground

These models run on WebGPU in your browser and score the options you supply in one forward pass. They return probabilities, not generated text. Requests and responses follow the TypeSafe Jev or OpenAI Decisions API shape, for every model. The first run of each model downloads its weights; the dropdown shows the download size and a Decision Index score, sorted highest first as a rough capability guide (higher is better; "proxy" means a related checkpoint's score).

Ready.

{}