SupraLabs

Try Supra-50M

51.8M parameters, running on your machine.

Params51.8M

QuantQ8_0

RuntimeWebAssembly

Weights stream straight from Hugging Face and inference runs entirely in this tab. Nothing you type leaves your machine.

Not loaded. The first load downloads 53 MB and is cached for next time.

Model ready. Ask it something.

Supra-50M is an Alpaca-style instruct model with a 1,024-token window, so each message is answered on its own rather than as a running conversation.

Weights / Supra-50M-Instruct on Hugging Face ↗