WebGPU
PulseAugur coverage of WebGPU — every cluster mentioning WebGPU across labs, papers, and developer communities, ranked by signal.
- developed by World Wide Web Consortium 100%
- founded by Apple Inc. 100%
- uses ONNX Runtime Web 90%
- uses WebAssembly 80%
- used by WebAssembly 70%
- used by ONNX Runtime Web 70%
- used by WebGPU Shading Language 70%
- used by Vulkan 70%
- used by Linux 70%
- uses Javascript 70%
- used by Javascript 70%
- used by PrismML 70%
22 day(s) with sentiment data
-
KetchUp: Open-source SketchUp alternative built with Rust and WebGPU
KetchUp is a new open-source alternative to SketchUp, developed using Rust and WebGPU. The project's creators view the open-source movement as a form of telemetry optimization, suggesting that users are inadvertently tr…
-
Local LLMs now run in browsers via WebGPU, enhancing privacy
The WebGPU API, now stable across major browsers like Chrome, Edge, and Firefox, enables the execution of local Large Language Models (LLMs) directly within the browser. This advancement, coupled with optimizations in O…
-
NKTg AI introduces deterministic text decoding to replace RAG before LLMs
NKTg AI has developed a new deterministic text decoding system designed to improve upon traditional Retrieval-Augmented Generation (RAG) methods. Unlike standard RAG, which uses vector embeddings and cosine similarity, …
-
llama.cpp enhances WebGPU support with new operations
The llama.cpp project has released updates focusing on WebGPU support, with recent commits adding functionality for COUNT_EQUAL and MEAN operations. These updates, assisted by Codex, aim to enhance the performance and c…
-
LFM3-Omni-2B model runs offline on iPhone 18 Pro Max
A user has successfully run the LFM3-Omni-2B language model offline on an iPhone 18 Pro Max, achieving a speed of approximately 80 tokens per second. This was accomplished using custom WebGPU kernels integrated into a R…
-
Weight-Space Programming: Edit LLM Weights Directly in Browser
A new technique called weight-space programming allows users to directly edit the numerical weights of language models within a web browser, bypassing traditional training methods. This approach, demonstrated by the web…
-
Ava: In-browser AI avatar runs locally with LFM3-Omni and WebGPU
Ava is a new in-browser AI avatar that can be run locally using WebGPU. It leverages the lfm3-omni model for its capabilities. The project aims to provide a more accessible and integrated AI experience directly within t…
-
AI agent apps run entirely in-browser using Laya decision model
A developer has created four AI agent applications that run entirely within a web browser using Laya, a 421M-parameter decision model from ConvAI Innovations, and WebGPU. These applications require no server or API keys…
-
llama.cpp project releases multiple pre-release updates with performance improvements
The llama.cpp project has released several pre-release versions, including b12121, b12119, b12118, b12116, b12114, b12110, b12109, b12102, b12100, and b12109. These updates address various issues and introduce new featu…
-
LFM3-Omni-2B model runs locally in-browser via WebGPU
The LFM3-Omni-2B model can now run entirely within a web browser using WebGPU, eliminating the need for separate audio and visual processing models. This advancement allows for local, end-to-end processing of visual, au…
-
Claude AI helps build PlayStation 2 emulator for web browsers
Developers have created TabStationToo, a PlayStation 2 emulator that runs entirely within a web browser. The project leverages the PCSX2 emulator's core and utilizes Claude for significant portions of the programming, i…
-
Shaders ships WebGPU effects as framework components with CLI and AI tooling
Shaders has released a new framework that allows developers to integrate over 200 WebGPU effects as components into their projects. The framework supports popular JavaScript frameworks like React, Vue, Svelte, and Solid…
-
LLMs in the browser face complex technical hurdles
Running large language models (LLMs) directly within web browsers presents significant technical challenges due to the complexity of browser environments and evolving web standards. Developers must navigate various comp…
-
LFM3-VL-2B model enables in-browser real-time video captioning
A new multimodal AI model, LFM3-VL-2B, has been released and can run in-browser using WebGPU for real-time video captioning. The model is available on Hugging Face, with a live demo showcasing its capabilities. This dev…
-
Slides released for talk on running LLMs locally in the browser
Thomas Bley presented slides from his talk "Run LLMs in the Browser" at iJS Munich. The presentation, available on GitHub, covers running large language models locally within a web browser. Topics include using tools li…
-
Qwen3.5-4B neuron steering playground launched for local browser use
A new interactive playground allows users to explore how specific neurons within the Qwen3.5-4B language model can steer its behavior. By manipulating small subsets of neurons, users can influence the model to adopt var…
-
Old CNN outperforms GPT-6 and Claude Fable 5.1 in skin cancer diagnosis
A 2015-era Convolutional Neural Network (CNN) has outperformed advanced large language models like GPT-6 Astra and Claude Fable 5.1 in diagnosing skin cancer from clinical and dermoscopic images. While the LLMs achieved…
-
Browser LLMs face load time and quality hurdles despite WebGPU advances
A developer explored running a large language model (LLM) for a real-time writing coach directly within a web browser using WebGPU. While a quantized 3.8B parameter model like Phi-3-mini was feasible, significant challe…
-
llama.cpp adds WebGPU support for Q2_0 quantization, boosting performance
A pull request to the llama.cpp project introduces support for the Q2_0 quantization format via WebGPU. This enhancement significantly boosts performance, with reported speedups of over 50x for certain models like Terna…
-
LLM copying improvements, llama.cpp WebGPU update, and human data validation needs
A new research paper proposes a novel method using VO-rotations and the Chinese Remainder Theorem to enable transformers to accurately copy repeated character sequences, addressing a known limitation in current frontier…