Try it

Write an email with a tiny language model running on your CPU, in this tab. Nothing is sent anywhere.

Expect mistakes. These models are a speed experiment: small enough to live in a CPU cache, so they are far less capable than a chatbot (why). Keep requests short and simple: “tell X …”, “thank X for …”, “ask X to …”.

Model

256 wide × 6 layers · 1.93 MB

Threads (single-threaded: page not cross-origin isolated) Creativity (temperature): 0.7 Signed as

new seed each time type it out

loading 4M (1.93 MB: download, compile, warm-up)…