Organizations shaping AI

AMI Labs

Advanced Machine Intelligence Labs is Yann LeCun's AI startup focused on world models and alternatives to token-prediction-centric language models.

Reka AI

Reka AI is a San Francisco AI research lab building models and infrastructure for physical AI. Founded by former DeepMind, Google, and Meta researchers, it joined forces with Moonvalley in June 2026, adding video-generation and world-modeling researchers to its team.

Tencent

Tencent develops the Hy foundation-model family for reasoning, coding, agents, and multimodal generation. Its broader AI program includes open image, video, 3D, and world models, delivered through Tencent Cloud and integrated across products such as WeChat, Yuanbao, Tencent Meeting, and its developer tools.

General Intuition

Frontier AI research lab spun out of the gameplay-clip platform Medal. Trains agents on large volumes of game video to develop spatial-temporal reasoning, with a $134M seed round.

Moonlake AI

Builds multimodal world models and simulation infrastructure for physical AI, generating structured, editable 3D environments with geometry, articulation, and physics for robotics, digital twins, and embodied-agent training.

World Labs

Startup founded in 2024 by Fei-Fei Li building 'large world models' that understand and generate 3D spatial environments.

Moonvalley

Generative-video company behind Marey, a production-focused model trained on licensed creative work. In June 2026, Moonvalley joined forces with Reka AI, bringing its video-generation and world-modeling research team into Reka.

Odyssey

Odyssey develops interactive visual world models: systems that generate video environments viewers can explore and influence in real time. The company is aimed at film, games and simulation, where controllable moving worlds matter more than producing a single finished image or clip.

Sooth Labs

Develops a world model for long-horizon global foresight, combining discrete events and continuous signals into calibrated, auditable forecasts for geopolitical, economic, and other real-world events.

Wayve

Pioneers an end-to-end approach to automated driving in which models learn directly from diverse driving data rather than relying on detailed maps and hand-built modules. Its research also spans driving-focused world models, vision-language systems and simulation.

Archetype AI

Physical-AI company founded by Ivan Poupyrev and colleagues, many formerly of Google. Builds Newton, a foundation model that fuses real-time sensor data with natural language to perceive and reason about the physical world.

Decart

Develops low-latency generative systems spanning model optimization, live video transformation and interactive world models. Its technology supports media applications and simulation for robotics, autonomous vehicles and other physical systems.

Skywork AI

Develops a cloud-native workspace and agent platform that combines deep research, multimodal content creation, and task execution across documents, slides, spreadsheets, websites, images, video, and other professional workflows. Its SkyClaw agent adds persistent memory, scheduled tasks, tool use, and messaging integrations.

Unreasonable Labs

Builds systems for scientific and technical discovery that connect knowledge across disciplines, generate and test hypotheses, and combine world models with simulations, tools, and human expertise.

AI lab building world-model systems for autonomous agents and robotics.