Kunlun Tech open-sources Matrix-Game 4.0, 60 fps on one RTX 5090
Skywork AI, part of Kunlun Tech, has open-sourced Matrix-Game 4.0: a game engine handles world state while a video model renders frames at 480p and 60 fps on a single RTX 5090.
World models News and Analysis
17 verified stories covering World models, product updates and industry developments.
AMD buys Fei-Fei Li's World Labs for $8.2 billion
AMD is acquiring Fei-Fei Li's World Labs for about $8.2 billion in an all-stock deal, with Li becoming AMD's chief scientist.
World Labs launches Atlas, a controllable 1440p video model
World Labs' new Atlas model bakes camera pose into its conditioning inputs, unifying 3D reconstruction, camera-controlled video generation, and spatial simulation in one architecture.
Runway's Solaris Renders Interfaces Frame by Frame
Runway's new Solaris renders app interfaces live, frame by frame, but 720p resolution and unstable text rendering keep it far from replacing real software.
Galbot lets robots learn from human video
Galbot, Peking University, CASIA and Tsinghua propose WAM-TTT, a deployment-time method that adapts robot behavior from ordinary first-person human videos without retraining the base world-action model.
Dexmal DW0.5 moves robot post-training into a learned virtual trial field
Dexmal DW0.5 moves robot post-training into a learned virtual trial field. The piece reviews the verified facts and why the signal matters beyond one announcement.
Being-M0.7 uses human-centered data to train humanoid motion
Being-M0.7 uses human-centered data to train humanoid motion. The piece reviews the verified facts and why the signal matters beyond one announcement.
PixVerse extends Series C funding while pushing AI video toward interactive games
PixVerse extends Series C funding while pushing AI video toward interactive games. The piece reviews the verified facts and why the signal matters beyond one announcement.
MoWorld runs a 50 FPS world model on Chinese NPUs
MoWorld runs a 50 FPS world model on Chinese NPUs is reframed for global technology readers, preserving the key numbers, claims and open questions.
Qwen-AgentWorld trains agents with a language world model
Alibaba’s Qwen team introduced Qwen-AgentWorld, a language world model that predicts how environments respond to agent actions.
Microsoft Mirage gives video models spatial memory
Mirage stores latent scene features instead of rendered point clouds, making video world-model generation faster and lighter while improving consistency.
Vast Unveils Project Eden, a Persistent AI World Model
Beijing-based Vast wants Project Eden to move AI 3D generation from single assets to shared, persistent virtual worlds.
Real-Time World Model Startup Raises $59M, Near-Zero Frame Latency
Reactor, a stealth startup founded by ex-Apple engineers, has raised $59 million in Series A funding to build a real-time world model platform that generates interactive video with near-zero latency.
Google turns Street View into a Genie world model
DeepMind connects 280 billion Street View images to Genie, making real places walkable as stylized 720p worlds for Ultra users.
Israeli AI Startup Decart Raises $300M, Backed by Karpathy and Nvidia
Decart announced a $300 million funding round at a $4 billion valuation, with participation from Nvidia and angel investor Andrej Karpathy.
Runway bets on world models, $5.3B valuation takes on Google
Runway, once known as an AI video startup, is now valued at $5.3 billion and pivoting to world models that understand physics from video data.
Germany's Sereact Raises $110M to Put a World Model Inside a Robotic Brain
Sereact, a German robotics company, has raised $110 million in Series B funding to scale its Cortex 2.0 "robotic brain" for industrial arms, using a world model and over a billion real-world picks to achieve a human intervention rate of once per 53,000 picks.