Bowen Wang

🎮 Computer Use Agent Arena is LIVE! 🚀 🔥 Easiest way to test computer-use agents in the wild without any setup 🌟 Compare top VLMs: OpenAI Operator, Claude 3.7, Gemini 2.5 Pro, Qwen 2.5 vl and more 🕹️ Test agents on 100+ real apps & webs with one-click config 🔒 Safe & free…

Big congrats to @TsingYoga and their team for pushing the boundaries of CUAs! When developing, UI-TARS-1.5 truly feels like the beginning of a new chapter — the next episode is coming. Stay tuned for the leaderboard🚀!

Meet Kimi-Researcher - an autonomous agent that excels at multi-turn search and reasoning. Powered by k 1.5 and trained with end-to-end agentic RL. Achieved 26.9% pass@1 on Humanity's Last Exam, 69% pass@1 on xbench. 🔗 Tech blog：moonshotai.github.io/Kimi-Researche…

🔥New Computer Agent Arena Leaderboard Updates (2k+ user votes)! 🤔Which VLMs act better as computer use agents (CUAs)? 1, Claude Sonnet 4 🥇 2, Claude 3.7 Sonnet 🥈 3, UI-TARS-1.5 🥉 4, Operator More insights in the thread 👇 arena.xlang.ai

Operator 🤝 OpenAI o3 Operator in ChatGPT has been updated with our latest reasoning model. operator.chatgpt.com

Based on my own testing, Claude 4 is even stronger in CUA than Claude 3.7 Sonnet, with agentic capabilities enhanced, come on and give it a try!

🤔Static CUA benchmarks enable fast model dev but lack task variety and risk overfitting. Computer Agent Arena tests crowdsourced real-world tasks. OSWorld: 🥇UI-Tars1.5🥈Operator🥉Claude 3.7 CUA Arena: 🥇Claude 3.7🥈Operator🥉UI-Tars1.5 🚀Rankings likely to evolve quickly

😀Our initial leaderboard finally came out, here I'd like to share a few interesting findings based on our case study: 1, Claude 3.7 Sonnet consistently performs best across diverse task types, particularly excelling at open-ended queries like “write a paper reading report.” 2,…

Part 2 of Build Your Own Operator on macOS is now live! The new cua-agent framework cuts down complexity and accelerates CUA development - so you can focus on building, not boilerplate.

For folks working on CUAs, definitely give o3 and o4-mini a try from @OpenAI. Key takeaway: Enhancing image reasoning and tool-use abilities on FM could significantly boost CUA performances.

UI-TARS-1-5

🚀 Meet Kimi-VL and Kimi-VL-Thinking! 🌟 Our latest open source lightweight yet powerful Vision-Language Model with reasoning capability. ✨ Key Highlights: 💡 An MoE VLM and an MoE Reasoning VLM with only ~3B activated parameters 🧠 Strong multimodal reasoning (36.8% on…

I want to highlight that this was an incredibly complex piece of work put together by @BowenWangNLP. We have been working on this for more than a year - much longer than a typical 3-5 month AI sprint. Big thanks to @taoyds for leading such an impactful project.