Why it matters

WebGPU enables browser LLM. Understanding shapes web deployment.

Advertisement

The architecture

WebGPU compute shaders.

WebLLM runs Llama / Mistral / Phi in browser.

Transformers.js higher-level.

WebGPU stackWebGPU APIbrowser GPU computeWebLLMruns LLM in browserTransformers.jshigher-levelChrome + Firefox stable; Safari growing; enables local LLM without install
WebGPU.
Advertisement

How it works end to end

Chrome stable; Firefox stable; Safari growing.

WebLLM: MLC-LLM WASM + WebGPU.

Transformers.js: HF port.