Releases·Tool Release
Software Mansion releases runntime, a TypeGPU-based framework for running AI models on the user's GPU in WebGPU environments
Software Mansion released ruNNtime, a framework for running AI models on the user's own GPU via WebGPU, built on TypeGPU. Kernels are written in TypeGPU and compiled to WGSL at runtime, so no WASM binary is needed and the same code runs in the browser, React Native, and Electron. It offers a 'zoo' of one-function-per-task models for speech, vision, and NLP, positioning itself as an alternative to transformers.js for on-device web inference.
37 engagement1 sourceMon, Sep 14, 2026, 01:28 PM
AI Overview
- ruNNtime runs AI models on the user's GPU via TypeGPU kernels compiled to WGSL at runtime, requiring no WASM download
- The `runntime/zoo` entry point offers single-function models for speech, vision, and NLP tasks, e.g. `createObjectDetector()`
- The kernels run anywhere WebGPU does: the browser, React Native, and Electron; `runntime/core` mirrors PyTorch semantics but is not yet public
- Software Mansion created runntime, released on GitHub September 14, 2026
Related
Products·launched
Browser-based image-text photo search using EmbeddingGemma 2 via ruNNtime (WebGPU)
A demo photo gallery where users search images by text description, with embedding inference running fully in-browser on WebGPU. It uses EmbeddingGemma 2's text and vision towers ported to ruNNtime, a TypeScript WebGPU inference library, mapping images and text into one shared 768-dim embedding space. It solves on-device, private image search without a server, targeting developers exploring browser-side multimodal inference.
A demo photo gallery where users search images by text description, with embedding inference running fully in-browser on WebGPU. It uses EmbeddingGemma 2's text and vision towers ported to ruNNtime, a TypeScript WebGPU inference library, mapping images and text into one shared 768-dim embedding space. It solves on-device, private image search without a server, targeting developers exploring browser-side multimodal inference.
79 engagement1 sourceredditWed, Oct 7, 2026, 09:26 AM
Releases·Tool Release
Show HN: A TypeScript PyTorch-style runtime that trains and runs inference in-browser on WebGPU
A developer launched a browser-based runtime that executes PyTorch-style code via WebGPU, with no installation, server, or account. The TypeScript-based tool supports both model training and inference and was built to let students run torch code in classes without setup.
A developer launched a browser-based runtime that executes PyTorch-style code via WebGPU, with no installation, server, or account. The TypeScript-based tool supports both model training and inference and was built to let students run torch code in classes without setup.
1 engagement1 sourcehackernewsWed, Sep 30, 2026, 04:15 AM
Releases·Tool Release
Hugging Face launches Fleet, browser-based WebGPU benchmarking with open-sourced kernels
Hugging Face released Fleet, a browser tool that runs WebGPU compute kernels from real AI workloads on users' GPUs and returns a personalized benchmark card. It also open-sourced hundreds of the kernels as a first step toward faster browser inference.
Hugging Face released Fleet, a browser tool that runs WebGPU compute kernels from real AI workloads on users' GPUs and returns a personalized benchmark card. It also open-sourced hundreds of the kernels as a first step toward faster browser inference.
25 engagement2 sourcesreddit, rssTue, Sep 1, 2026, 04:00 PM
Releases·Tool Release
browser-llm-fit: Check if an AI model fits the browser
A developer released browser-llm-fit, a JavaScript library that probes a client's WebGPU hardware limits — such as maxStorageBufferBindingSize and shader-f16 support — and ranks which AI models can run in the browser before weights are downloaded. It targets crashes from oversized or unsupported model loads in WebGPU tabs.
A developer released browser-llm-fit, a JavaScript library that probes a client's WebGPU hardware limits — such as maxStorageBufferBindingSize and shader-f16 support — and ranks which AI models can run in the browser before weights are downloaded. It targets crashes from oversized or unsupported model loads in WebGPU tabs.
1 engagement2 sourcesreddit, hackernewsThu, Sep 3, 2026, 04:59 PM
Releases·Tool Release
Hugging Face open-sources WebGPU kernel collection for fully local browser AI, plans upstreaming to Transformers.js and ONNX Runtime Web
Hugging Face released open-source WebGPU kernels claimed to be the fastest for local AI, covering 200+ common ML operations that run entirely in the browser. The team plans to upstream the optimizations into Transformers.js, ONNX Runtime Web, and LiteRT.js, making fast local inference more accessible to web developers.
Hugging Face released open-source WebGPU kernels claimed to be the fastest for local AI, covering 200+ common ML operations that run entirely in the browser. The team plans to upstream the optimizations into Transformers.js, ONNX Runtime Web, and LiteRT.js, making fast local inference more accessible to web developers.
624 engagement1 sourceredditWed, Sep 30, 2026, 04:02 PM