webgpu

A 1.6 MB model using ternary 3:4 weights runs on WebGPU in the browser, playing Connect Four as well as a 7.8 MB int8 model, demonstrating the effectiveness of ternary formats for lightweight AI deployment.

The article explains subtle differences in struct layouts between Rust and WebGPU, and introduces a unit testing approach using naga to verify memory offsets for safer GPU compute shader configuration.

A collaboration between TresJS and PMNDRS to release official support for a typography engine for web graphics using WebGPU.

The article discusses the integration of compute shaders into p5.js to simplify teaching GPU programming through scaffolded learning, addressing educational challenges in computer graphics.

A new technique called JEVfire enables existing LLMs like Qwen to behave more like Jev by modifying decision-making processes without retraining, resulting in significantly faster JSON generation and enabling local AI agents to run efficiently on consumer hardware.

vgpu is a TypeScript library for WebGPU, featuring typed shader imports, a tiny GPU-first API, and compatibility across browser, headless Node, and test environments.

AIRI is an AI companion that features real-time voice chat, gaze-tracking avatars, memory retention, and gameplay support for Minecraft and Factorio, with compatibility for over 30 LLMs and local inference via WebGPU.

Ternary Bonsai 2 is a 27B parameter model derived from Qwen3.8-27B that uses ternary weights to achieve a size under 6GB while retaining 98.2% of its intelligence, enabling it to run in-browser on WebGPU.

A coding agent that runs entirely in the browser using the Pi framework and MiniCPM5-2B AI model with webGPU acceleration, showcased on Hugging Face Spaces.

A WebGPU shader on an untrusted site can freeze a Mac's graphics, making the desktop UI unusable until a forced restart, with similar issues previously addressed by Apple.

Mentria.ai's browser-based inference engine achieves 25-30 tokens per second with a 1-bit 27B parameter model on a 6 GB RTX 3060 laptop GPU using WebGPU, enabling efficient AI inference without server installation.

gpu-lexer is a WebGPU-based, language-agnostic syntax highlighter that uses a tiny AI model to label code parts without predefined grammars, offering a novel, performance-focused alternative to traditional highlighters.

A tiny 27.5KB AI model called gpu-lexer from Vercel Labs runs locally in the browser using WebGPU to syntax-highlight code in any programming language without prior knowledge.

WebLLM is a high-performance in-browser LLM inference engine that leverages WebGPU for hardware acceleration and is fully compatible with the OpenAI API, enabling local execution of open-source language models.

Hugging Face releases 207 WebGPU kernels for local AI inference in browsers, along with a JavaScript loader library and a benchmarking tool called Fleet.

A tweet highlighting the potential of WebGPU for building impressive interactive demos, featuring a playful challenge about a cube falling into a hole.

Vercel has open-sourced vgpu.sh, an agent-native development tool for WebGPU that enables shader execution in browsers and headless Node.js, with features for CI testing and agent integration.

Spline V2 is a completely rebuilt 3D editor platform for the agentic era, introducing AI Agent Mode, WebGPU engine, and custom scripting features.

Galaxium is a tool for exploring the universe using WebGPU technology, offering interactive and graphical experiences.

Pngine is a declarative format and runtime for WebGPU that uses S-expressions to simplify declaring and validating WebGPU configurations, enabling cross-platform sharing and export to formats like HTML or PNG with embedded runtime.