Sage 2

A little browser model with a built-in knowledge base. Gemma 4 E2B trunk (int4) plus Sage's seam, all in WebGPU. Every token generated on your GPU — no server, no API key, no upload.

uninit Idle · load the model to begin.

Load the model

First load streams the 1.5 GB Sage 2 artifact and dequantizes it to F16 on the GPU (about a minute; a quarter of that from the browser's HTTP cache). After that every answer is generated locally. Ask it what its knowledge base records about something.

Requires a WebGPU browser with the shader-f16 feature.