Introducing GLiNER2.5-Decide, our new 340M parameter open weight, encoder-based decision model. GLiNER2.5-Decide is built for fast, deterministic classification. The model evaluates a set of user-defined typed questions and rules, and jointly decodes their answers, returning structured decisions with probability distributions and confidence scores. We evaluated the model’s performance on Fast Decisions, an unseen, internally generated classification suite based on 17 datasets testing real-world use cases across routing, triage, classification, sentiment, and content understanding. Measured against similar decision models, GLiNER2.5-Decide leads in 9 of the 17 datasets, achieving the highest average score: - GLiNER2.5-Decide: 60.1% - SemIf: 56.4% - JevK5: 57.5% - Laya: 46.6% This performance makes the model a strong fit for use cases like tool calling, model routing, browser and computer use, and LLM-as-a-judge. GLiNER2.5-Decide’s lightweight encoder architecture makes it easy to fine-tune the model for specific tasks, while being efficient enough to run locally on consumer-grade CPUs or in air-gapped environments, giving users greater control over where their data goes and where the model runs. To make building and experimenting with GLiNER2.5-Decide as easy as possible, we're also offering hosted inference. You can now use our API to run inference and fine-tune GLiNER models on specialized tasks right inside your own coding agent: agent.fastino.ai As with previous models, we’re also releasing the model weights on @huggingface under the Apache 2.0 license: huggingface.co/fastino/GLiNE…

NEW: Ternary Bonsai 2 just landed on Hugging Face — a 27B reasoning model in ternary weights, built for the agentic era 🤯 9x smaller than FP16 while retaining 98.2% of the intelligence. At <6GB in size, it can even run 100% locally in your browser on WebGPU! Try it yourself 👇

Some cool new packages we've made! @⁠remotion/whisper-webgpu: Transcribe audio in the browser @⁠remotion/gsap: Use GSAP to animate in Remotion @⁠remotion/mac-cursors: Every Mac cursor as vector graphic Happy hacking 🥳 remotion.dev/api

How much AI can you run in the browser today? @nicodotdev from @huggingface joins @grabbou to explain how Transformers.js brings pretrained models into JavaScript for text, audio, image, and agent workflows. The conversation covers ONNX Runtime, WebGPU, model downloads and browser caching, CPU fallbacks, performance across devices, local speech recognition, and background removal. They also discuss browser agents and Hugging Face's early WebGPU inference engine, where current experiments show 5x to 10x speedups. Chapters: 0:43 Welcome to React Universe On Air 1:04 Nico Martin's work at Hugging Face 2:04 What Transformers.js is 3:30 Browser AI beyond large language models 6:00 Bringing Python ideas into a JavaScript API 10:08 How the Transformers.js pipeline API works 14:23 Choosing compatible models and architectures 15:46 Why Transformers.js uses ONNX Runtime 19:01 CPU inference, WebGPU, and browser coverage 22:57 Model downloads, browser caching, and UX 28:28 Startup time and performance across devices 30:57 Why Transformers.js focuses on the browser 34:07 What developers misunderstand about browser AI 38:44 Building AI agents in the browser 43:19 Local fallbacks and hybrid AI 44:28 The browser AI roadmap 47:04 Hugging Face's custom WebGPU inference engine 50:57 Why speedups matter on constrained hardware 52:07 Nico's favorite browser AI use cases

Background removal and transcription run in the browser. @nicodotdev joins @grabbou to cover Transformers.js today, device limits, offline use, and which parts of an agent workflow can already run locally. 📺 YouTube premiere: piped.video/elAVSra9KCY 📆 Sept 7 at 2PM CEST