LingBot-World 2.0 made its stunning debut at the IFA Berlin opening keynote! @jackhuynh from @AMD demonstrated LingBot-World 2.0 live and praised it as “the future of Personal AI.” With just a single prompt, the LingBot-World 2.0 generated a vivid, interactive 3D world—from traditional Chinese towns to European countryside. 🌍 You can freely roam, observe, move, and interact with the dynamically expanding environment in a long-term, stable, real-time simulation. #LingBotWorld2 #IFA2026 #PersonalAI #EmbodiedAI #WorldModel

SGLang v0.5.18 is out! We have 710 PRs from 212 contributors, and 61 of them were new contributors. Welcome🧡 Highlights: 🔸 New models: Muse Glimmer and Intern-S2-Mobius, plus SANA-Video, LTX-2.5, LingBot-Video, Cosmos3, and LongCat-Image on SGLang-Diffusion 🔸 Engine startup is up to 2.38x faster: checkpoints stream from disk while CUDA graphs capture, so the two longest waits happen at once 🔸 All compiled kernels (Triton, FlashInfer, Inductor, DeepGEMM) now live in one directory: SGLANG_CACHE_DIR 🔸 NVFP4 checkpoints now run on AMD GPUs via online MXFP4 requantization, with 97.5-100% accuracy retention across MiniMax-M2.7, GLM-5.1, Kimi-K2.6, Qwen3.5-397B, and DeepSeek-R1 🔸 Kimi K3 runs much faster on AMD MI355X: 1.37-1.77x throughput and up to 2.42x faster ITL 🔸 DeepSeek-V4 decode gets faster on NVIDIA: TP LMHead all-to-all cuts LMHead time from 320us to 169us, and FlashInfer MNNVL allreduce turns on automatically for DeepSeek models Thanks again to our amazing partners and model makers: @NVIDIAAI @AIatAMD @intel @AIatMeta @intern_lm @Lightricks @AntLingAGI @robbyant_brain @Meituan_LongCat Full release note👇

👋 WRC 2026 is a wrap. 🤖 Thank you to everyone who visited Robbyant booth. We showcased LingBot 2.0 alongside real‑world deployments across smart pharmacy, logistics sorting and industrial machine‑loading scenarios. 🧠 One model, running on 3 distinct robot embodiments for 3 different real‑world tasks. 📍Next stop: ECCV. See you there!

Robbyant R2 has been recognized by IDEA and K-Design Award. 🏅 For humanoid service robots, great design is not just how it looks — it is how safely, efficiently, and naturally it works around people. Our R2 industry solution will be showcased at WRC this week. Come meet us at C111.

Some Theoretical and Practical Thoughts on Diffusion Language Models jzhao2024.github.io/notes/20…

Some Theoretical and Practical Thoughts on Diffusion Language Models jzhao2024.github.io/notes/20…

This week in Shanghai, PyTorch Foundation Executive Director @sparkycollier and PyTorch Foundation CTO @matthew_d_white will speak across three forums at World Artificial Intelligence Conference 2026. On July 18, Mark will participate in @alibaba_cloud's “Agentic Cloud: The Infrastructure for the Agent Era” forum (Silver Hall C+D, Layer G, Shanghai World Expo Center), when Mark will participate in the 11:40 dialog “Build trusted agents at scale with Agentic Cloud” with Ali AI Group CTO Feifei Li. On July 19, Mark and Matt will deliver keynote talks at BAAI’s Open Compute for AI and AI Agent Ecosystem Forum (Room 618, Shanghai World Expo Center). Their sessions will address how open source is rewiring the future of AI and technical autonomy across hardware architectures, supply chains, and agentic systems. Later that day, Mark and Matt will deliver keynote talks at @AntGroup's Making large models accessible: AGI Infrastructure Forum (Room 518, Shanghai World Expo Center). Mark will present “The Open Acceleration: How Open Source is Rewiring the Future of AI” at 14:05, followed by Matt’s “Just Enough Intelligence: Why the Key to Widespread Adoption of Intelligence is Optimization Engineering” at 14:20. Mark and Matt will then participate in the 14:35 dialogue “Open Foundations, Inclusive Futures: Collaborative Infrastructure for Accessible AGI” with Ant Group CTO Zhengyu He. 📸 Check out Mark exploring the conference and speaking with CCTV during an on-site interview. Stay tuned for more from Shanghai.

🤖 LingBot-VLA 2.0 is now open-source — our next-gen embodied foundation model. 🔷 60,000 hours of high-quality pretraining data — combining curated robotic demonstrations and egocentric human operation videos 🔷 20 robot configurations across 17 brands — Astribot, Leju, Unitree, Franka, Fourier, Realman, and more 🔷 Whole-body DoF: heads, waists, dexterous hands, and mobile bases — enabling far more complex task scenarios 🔷 Inference under 130ms on RTX 4090 — developer events launching soon #EmbodiedAI #Robotics #OpenSource #VLA

Today we open-source LingBot-Video — the first MoE-based video foundation model built for embodied intelligence. 🔹30B params, only 3B active at inference. 🔹Augmented with 70K hours of embodied data on top of large-scale internet video pretraining. 🔹Already outperforming Wan2.6, Seedance 1.5 Pro, and Cosmos3 Super on RBench. 🧵👇

See the depth prediction for robot manipulation on the most challenging case: TRANSPARENT objects! A plastic bag pouring beans into glass jars, and the perceived depth stays consistent over time, with clean and sharp BOUNDARIES. Why is the depth this good? The hero behind the LingBot-Depth 2.0 upgrade is LingBot-Vision, our new self-supervised vision pretraining that makes boundary geometry the learning signal of SSL. The model discovers its own boundary tokens from raw images (no labels, no edge detector, bottom-left panel) and forces exactly those into the mask, so the student must reconstruct boundary geometry from context. Bottom-right: a single frozen-feature query staying locked on the deforming bag.

Fully Autonomous Operation. What's driving them? The answer starts tomorrow. 🔥 #LingBot2 #Robotics #EmbodiedAI

World models feel like the future... almost... We can still see some weird artifacts. That means we need high-quality benchmarks and the folks at @Meituan_LongCat know that. WBench, a high-quality world generation benchmark for Interactive Video World Model Evaluation 😍 Benchmarking: 1. video quality 2. Consistency 3. Physics compliance 4. Whether the model generates what it's told (important) 🏆 Current top contenders 🥇 @robbyant_brain 🥈 @TencentHunyuan 🥉 @TencentHunyuan I can't wait for models to generate real-time, super-immersive worlds.