Browser language model

An experiment with a 30-million-parameter language model. Load it on your device and try a short conversation.

Stentor-30M-Instruct

30M parameters. 512-token context. Runs on this device.

The first load downloads the model. Your prompts stay in this browser.

When the model is ready, send it a short question.

Model by StentorLabs, quantized to GGUF Q4_K_M by mradermacher (Apache-2.0)