PrismML Ternary Bonsai 1.7B — every weight is −1, 0, or +1 — executing on your GPU via WebGPU. The exact same pure-Rust code (Ferric) runs this natively and here in WASM. Nothing is sent to a server; the model downloads once and inference is local.
Click Generate — first run downloads the ~442 MB model (cached after).