Three-LLM, a WebGPU-based LLM inference engine utilizing Three.js, has been released.
Three-LLM: A WebGPU-based LLM Inference Engine Using Three.js Released
This article is a translation. Read the Japanese original
This project aims to enable LLM inference directly within the browser by leveraging WebGPU.
Three.js is used for the implementation.
Technical details are reportedly available on the official website (https://three-llm.ben3d.ca).
Source: Three-LLM: Three.js-based WebGPU LLM inference engine (Hacker News Frontpage, 2026-09-04)