Google expands Project Genie with Street View-grounded worlds
Google added Street View grounding to Project Genie and began rolling the experimental prototype out to eligible Google AI Ultra subscribers globally.
World model timeline
World model timeline for launches and milestones.
Use the world model timeline to follow dated launches, demos, model releases, and platform milestones.
LaunchesOpen sourceConsumer signalsPhysical AI
Milestone reelOpen signal
Runway turns GWM-1 into a real-time character product surface
Runway said its new Characters system is built on GWM-1 and made the real-time conversational video agent available through the Runway API plus web and mobile apps.
Ant Lingguang brings a world-model experience to mobile
Ant's Lingguang App was reported to add an 'Experience World Model' feature backed by LingBot-World-Fast, turning one image into a short explorable 3D scene on mobile.
Tencent open-sources HY-Embodied-0.5 and extends it with 0.5-X
Tencent open-sourced HY-Embodied-0.5 for real-world embodied agents and then released the HY-Embodied-0.5-X enhancement with inference and training code.
Alibaba launches HappyOyster for real-time world creation
Alibaba introduced HappyOyster as an open-ended world model product for creating, exploring, and directing immersive environments in real time.
Tencent Hunyuan opens HY-World 2.0 for persistent 3D worlds
Tencent Hunyuan expanded HY-World 2.0 beyond its April release by publishing HY-Pano 2.0 inference code and weights on May 11, then world-generation inference code plus WorldStereo 2.0 weights on May 18.
Robbyant releases LingBot-Map for streaming 3D reconstruction
Robbyant followed the LingBot-Map release with public evaluation scripts for KITTI and Oxford Spires, strengthening the model's reproducibility story for streaming 3D reconstruction.
NVIDIA previews Cosmos 3 as a unifying physical-AI world model
At GTC 2026, NVIDIA said Cosmos 3 would unify synthetic world generation, physical-AI reasoning, and action simulation, then kept widening the public Cosmos surface through an open data-factory blueprint and GitHub repos that were still active in late May.
Robbyant publishes LingBot-VA for robot-control world modeling
Robbyant released LingBot-VA as a causal video-action world model that predicts visual dynamics and robot actions together.
Robbyant releases LingBot-VLA for embodied-AI control
Robbyant released LingBot-VLA as a vision-language-action foundation model for generalist robot manipulation, with public code, paper, and model checkpoints.
World Labs announces the World API
World Labs moved its world model work toward a programmable developer surface with the World API.
Project Sid frames many-agent simulation as an AI civilization experiment
Project Sid gives the timeline a society-layer signal: agents specialize, coordinate, trade, and transmit behavior inside a persistent Minecraft-based world.
Runway introduces GWM-1
Runway framed GWM-1 as a general world model direction that reaches beyond video into worlds, avatars, and robotics.
World Labs launches Marble
Marble gave the world model category a concrete product surface: generated 3D worlds that can be explored and edited.
Google DeepMind announces Genie 3
Genie 3 placed interactive, navigable world generation near the center of the world model conversation.
NVIDIA launches Cosmos
NVIDIA introduced Cosmos as a world foundation model platform for physical AI development.
Oasis demonstrates a realtime playable AI world
Oasis demonstrated an action-conditioned Minecraft-like world that can be generated frame by frame while the player moves.
Skybox AI keeps the entry point simple: type a place, open a world
Skybox AI gives the timeline an early consumer-facing spatial signal: a generated world can start as a 360-degree horizon, mood, and environment shell.