Our new multi-model agentic security system brings together more than 100 specialized agents across frontier and custom models to find exploitable bugs, delivering top performance on the CyberGym benchmark. We used it ahead of Patch Tuesday to help find and fix 16 vulnerabilities. Today we’re announcing that customers can sign up to test it in private preview.

Supporting @MITEECS and @nlp_mit’s Multimodal Machine Learning course (Spring 2026). 🎓 Students are leveraging the multimodal capabilities of Kimi K2.5 to power their final research projects. We look forward to seeing the innovative applications that will emerge this semester. 🔗 Happy coding! ✨

499 Detalks [CN] 225: Deconstructing the Agentic Web: Multi-Model Intelligent Routing & On-Chain Financial Infrastructure Accelerating AGI Landing Time：28, April, 2026, 20:00 (UTC+8) Hosted by 499 and co-hosted by @BAI_AGI, this Space explores how intelligent model routing, on-chain data infrastructure, and dedicated agent financial rails can bridge the gap and accelerate AGI into the real world. Mod: Charis @charis_em, 499 Core Member Guest Speakers: Jtsong @Jtsong2, Head of APAC, @0G_labs Anita @Anitahityou, APAC, @SentientAGI Leslie @leslieloser_ , Content Creator Mia @Artistkatty_ , Growth from Starchild Rika @rayrayweb5, KOL Key discussion topics include: -- From “chat tools” to “autonomous economic entities”: What is the single biggest bottleneck stopping AI Agents from true mass adoption -- Multi-model intelligent routing & data orchestration: How far has the industry come in routing across multiple models and data sources -- Agent Wallets & on-chain credit systems: The power of Machine-to-Machine (M2M) value transfer when every AI has its own wallet Space Link: Telegram Group:

One of my favorite Grok features is his multimodal capabilities. Every morning I upload my sleep timeline screenshot to Grok. Every morning he tells me the truth: 5h 14m total, you need more sleep. Every morning I ignore him. I have an AI race to win. Builders, you in the same boat? @grok @xai

We're introducing WorldVQA, a new benchmark to measure atomic vision-centric world knowledge in Multimodal Large Language Models. Current evaluations often conflate visual knowledge retrieval with reasoning. In contrast, WorldVQA decouples these capabilities to strictly measure "what the model memorizes." The benchmark consists of 3,500 VQA pairs across 9 categories, with careful attention to linguistic and cultural diversity:

nemotron 3 nano omni-30B reasoning at Q8 running autonomously on my dgx spark right now. 58 tok/s. 1 million context. multimodal. hermes agent is using it to research xAI's new grok algorithm that dropped yesterday. pulling repos. scanning code. breaking it down. all while i post this. 30B model. 58 tokens per second. 1M context. reads images and video. locally. for free. nobody is talking about this model and that's insane.

Origin Lab $8M Seed Round⚡️ 📑 About: @OriginLabAI is a marketplace for premium, rights-cleared video game and 3D multimodal datasets for AI training. 🤝 Investors: @lightspeedvp (Lead), @svangel, @EniacVC, FPV Ventures, @SevenStars_VC, and @kevinlin

We're heading to AI Engineer Singapore this weekend (May 15–17)! 🇸🇬 Two things to know: 1️⃣ Vincent Wu, our Solutions Engineer, will be speaking on Building Agentic, Multimodal Systems with MiniMax. 2️⃣ We're co-hosting a side event with @zocomputer on May 15 evening. Come build with us → Thanks @aiDotEngineer @swyx @SherryYanJiang for making this happen!

Meet Qwen3.6-27B, our latest dense, open-source model, packing flagship-level coding power! What's new: Outstanding agentic coding Strong reasoning across text & multimodal tasks Supports thinking & non-thinking modes Apache 2.0 Smaller model. Bigger results.

📢 Kimi K2.6 API is live • Input Price (Cache Hit): $0.16 / M tokens • Input Price (Cache Miss): $0.95 / M tokens • Output: $4.00 / M tokens Kimi K2.6 is our latest + most intelligent model - stronger long-horizon coding, better instruction following & self-correction. Native multimodal (text/image/video), thinking + non-thinking, 256K context. Supports tool calls, JSON / Partial mode, and web search. Try it →