webgpu

WebLLM is a high-performance in-browser LLM inference engine that runs locally in the browser with WebGPU acceleration. It exposes an OpenAI-compatible API, supports streaming and JSON mode, and includes examples for building chat apps and browser extensions.

A 30-second WebGPU benchmark that measures a device’s memory-bandwidth ceiling for local AI and checks for a WebGPU subgroup correctness bug, with receipts, projections, and comparison data.

Sipp is a zero-dependency AI inference runtime for running models in the browser and across Node, Rust, Python, and self-hosted gateways. The page shows install commands, a symmetric client API, benchmark claims, and live demos for chat, agents, vision, and game use cases.

A browser-based procedural 3D world built by Claude Fable 5 with three.js and WebGPU. The repo describes a fully generated open world, its rendering and world systems, and how to run it locally.

A browser-based computer vision project that detects cursor hotspots in screen recordings using temporal motion cues, ONNX Runtime Web, and WebGPU. The repo includes a live demo, training scripts, synthetic data generation, and baseline comparisons against YOLO models.

A browser-based clicker game where local AI builds a custom game from your idea entirely on-device via WebGPU, with no API keys or servers.

A long-form technical guide on running local LLMs in the browser, with checks for WebGPU support, model hashes, tokenizer matching, and browser-side data leakage.

Browser-only local AI app for running compatible Hugging Face ONNX community models with WebGPU. The page shows a model library, local text generation, and an inference console in the browser.