Checking WebGPU

Tenzan TinyServe

Watch a language model think, live in your browser.

Seven views show what a small language model does at each token.

Checking if your browser can run the model (WebGPU)…

    Details

    Ask

    Load the model to ask your own question.

    Checking the size asks huggingface.co for the model's file sizes.

    Transcript

    Recorded run of 101 tokens on Apple Metal-3, played slower.

    01 Prefill vs Decode

    Models