Private by design

Run a language model on your own GPU.

Choose a model to download it once. It then runs entirely in this tab: a Rust engine compiled to WebAssembly computes every token with WebGPU shaders. Your messages never leave this device.

• No install, no account. A modern browser is all it takes.
• Stays on your device. No server sees your prompts or replies.
• Works offline once a model is saved in the browser.