Back to AIBriefs
LaunchDevelopersAI AgentsAI Models

Fable 5's WebGPU kernels run Gemma 4 at 255 tok/s in browser

Fable 5's WebGPU kernels run Gemma 4 at 255 tok/s in browserxcancel.com

Fable 5, an agentic optimizer, wrote custom WebGPU kernels achieving 255 tokens/s for Gemma 4 inference in-browser. The demo and kernels are now released on Hugging Face Spaces.

RT @osanseviero: Gemma 4 with 255 tokens per second running directly in your browser...4 days agoHugging Face
RT @xenovacom: Before Fable 5 was shut down, it pushed Gemma 4 to 255 tok/s on WebGPU. Some didn't...4 days agoHugging Face
Gemma 4 E2B running in-browser at 255 tok/s using WebGPU kernels written by Fable 54 days agoxenovatechDiscuss
··Discuss
Jun 18, 2:14 PM
Fable 5's WebGPU kernels run Gemma 4 at 255 tok/s in browser — AIBriefs