WebBrain

NVFP4 quantized Nanbeige 4.3 33B: huggingface.co/webbrain-one...

WebBrain Companion is live! Now you can use your existing subscriptions with WebBrain: www.webbrain.one/companion/

A lightweight 1B-parameter sufficiently good WebGPU model for text/vision and decisions in web-related tasks: huggingface.co/webbrain-one...

Security notice: We removed malicious code from WebBrain’s main branch today. Chrome and Firefox store releases were not affected. Devs who pulled the repo: please review the incident details. Current main is fixed. Investigation is ongoing. github.com/webbrain-one...

Gemini 4 Argon support is coming soon github.com/webbrain-one...

WebBrain 40 is live on GitHub and Mozilla AMO, and it's soon coming to the Chrome Web Store.

Found a great tutorial on WebBrain on YouTube: www.youtube.com/watch?v=Ilr7...

d1-omni (by @liquidai) is a compact multimodal decision model. We removed audio, adapted it for browser decisions, and exported it for local WebGPU inference. huggingface.co/webbrain-one... Coming to WebBrain 40 as an optional local decision provider, alongside many new features.

Apocalypse Mode explained

🤷‍♂️🥲

WebBrain v40 is coming with Steering! If you navigate to another site or interact with the page (changing the DOM) in the middle of an operation, it'll recover on its own. Big! And a first among browser agents, as far as I know.

SafeSocial is now live on Google Play. It uses an on-device AI model to blur comparison triggers on Instagram (romance, luxury, etc.), helping keep your feed intentional. www.webbrain.one/blog/safesoc...

We put OpenRouter's super inexpensive Nex N2.5 model offerings to test. At ~$0.15 for 100 agent turns, the price advantage is massive—but how does tool selection hold up against Sonnet and MiniMax M3? Results and benchmarks: www.webbrain.one/blog/nex-n25...

We just released WebBrain Compass Tiny XS v3 on Hugging Face: a compact 1.7B model fine-tuned for in-browser agentic execution, structured tool calling, and fast local decision making. Weights and LoRA adapter: huggingface.co/webbrain-one...

Introducing SafeSocial 🛡️ by WebBrain Your Instagram feed, now cleaner!

We just released WebBrain 36.8.0 with Jev and SafeSocial. Both use small models to assist a larger system. But our results were mixed—and taught us a great lesson on when small models actually work in production 🧵👇 www.webbrain.one/blog/webbrai...

testing jev on webbrain!

WebBrain's phone integration is unreal. It handled a call for me last week and saved our founder a ton of time. Watch 👇

Less than 2 months after surpassing 1B monthly tokens and 500 GitHub stars: pressreleases.online/pr/webbrain-... We've now surpassed 15B monthly tokens and 1,100 GitHub stars. The appetite for AI worldwide is mesmerizing indeed.

WebBrain 36.0.0 is coming with a lot of fixes. Check out the changelog: github.com/webbrain-one...

WebBrain-Compass-Tiny-v1 looks promising Qwen 3.6 35B-grade inference embedded in your browser via WebGPU! Runs at just 1/10th the total size (matching only the active params). Because the web has everything we need! It's coming 🔜 www.webbrain.one/blog/webbrai...

🎉 Milestone Alert! 🎉 We just passed 1,000 stars and 50 contributors on @GitHub! Huge thanks to everyone starring, contributing code, testing local models, and sharing feedback. WebBrain wouldn’t be here without you. 🤗 🔗 github.com/webbrain-one... #OpenSource #AI #LocalAI

There was a problem with our GitHub repo the other day. It’s fixed now, so you can continue starring responsibly until it reaches 1,000 stars. ⭐️ github.com/webbrain-one...

We’re nearly at 1,000 GitHub stargazers. Only 10 spots left on the WebBrain founding stargazers list 😊 Star responsibly: github.com/webbrain-one...

Try WebBrain-VL-2-450M in your browser. The HuggingFace open-source team built a live demo on Spaces for us 450M params, beats 4B-class models on our benchmark. huggingface.co/spaces/huggi...

WebBrain works with your existing AI subscriptions (ChatGPT, Claude, Gemini, Kimi). No extra accounts required. Here's how:

I'm testing WebBrain with a local proxy

WebBrain Cloud just got much faster!

100 forks on our GitHub repo! 🎉 And right on the heels of Google featuring WebBrain in the Chrome Web Store. It’s been a great week for WebBrain. Give it a try if you haven’t already—and definitely keep it around for Apocalypse Mode. 😄 webbrain.one

AI agents can now edit their own websites 🤯

WebBrain can save you money. Those pesky subscription pages getting in the way? Let WebBrain handle them and cancel the subscriptions you no longer need. webbrain.one

Introducing Apocalypse Mode — AI for when the internet disappears. webbrain.one

Qwen 3.8 27B + WebBrain, running fully offline, works like a charm. Give it a try if you have an NVIDIA 5090 or a MacBook with 48GB+ of memory. #Qwen #WebBrain #LocalAI #OfflineAI #AI

New release: Ling-3.0-tiny Q4 ONNX. Ant Group's tiny but capable model, quantized to Q4 ONNX, now runs fully offline in the browser as WebBrain's WebGPU fallback inference engine. huggingface.co/webbrain-one...

WebBrain is live on ProductHunt today 🚀 In a remarkably short time we’ve grown to thousands of users, billions of tokens served monthly, 700+ GitHub stars & 40+ contributors And happy to have @chrismessina.me hunting us today Would love your support & feedback www.producthunt.com/products/web...

Humbled that our number of Chrome-only active users surpassed 1,000 for the first time. github.com/webbrain-one... #Chrome #ChromeExtension #WebExtension #Milestone #OpenSource #GitHub #WebBrain #offlinellm #qwen #gemma #ollama #llamacpp

Who will be our 700th star? 😊 github.com/webbrain-one...

DeepSeek V4 Flash is no longer text-only 👀 In our internal benchmarks, it delivered significantly better price-performance than others in its class. We’ve added vision for screen-level understanding, capabilities needed for WebBrain Available on HF: huggingface.co/webbrain-one... #deepseek #ai

We tested 13 OpenRouter models across 1,300 browser-planning calls. DeepSeek V4 Flash led consensus, Gemini 3.6 Flash delivered 100/100 valid calls, and Kimi K3 agreed with Sonnet 5 on 86% of tool choices. Full benchmark: www.webbrain.one/blog/america...

🚀 WebBrain 26.0.2 is out — built with contributions from 6+ distinct contributors. Thanks to everyone who made it happen! #release #opensource #llm #ai

sending this with kimi k3 on webbrain

We just hit 500 stars on GitHub! Huge thanks to everyone for your interest and support. We hope you're enjoying the project—if you haven't already, please leave a star and help us spread the word! github.com/webbrain-one...

Yesterday was our biggest day ever: 100M tokens served on WebBrain's Cloud alone. 🤗 That's not counting users running their own keys or local models. Wild week.

The ultimate LLM won't be wrapped in an agent framework. The goal is an LLM that is agentic by nature. No wrappers, no external tool-managers—just native agency. #AI #LLM #AGI #AgenticAI #AIagents #AutonomousAgents #MachineLearning #ML #OpenSourceAI