~/local-llm-playground / 01_chat $
airgap_ok │ gpu: WebGPU

open ./01_chat.html

Basic chat with Gemma 3 270M running entirely in your browser via WebGPU. No server, no cloud, no data leaves your machine.

How to get the model file — the .task file must be downloaded once (~500 MB):
  1. Accept the Gemma license on Kaggle › gemma3-270m-it-int4
  2. Download gemma3-270m-it-q4_0-web.task
  3. Select it with the file picker below, then click Load Model
setup step 1 / 2 // select model file
Select a .task file to begin.
chat step 2 / 2 // ctrl+enter to generate