Generalist AI

Generalist AI is a frontier artificial intelligence and robotics company established in 2024 with the ambitious mission of bringing general-purpose robots to reality. The company is co-founded by a team with deep roots in the AI and robotics world: Pete Florence, who serves as CEO; Andrew Barry, the CTO; and Andy Zeng, the Chief Scientist. This founding team brings a wealth of experience from leading research labs and companies such as Google DeepMind, OpenAI, and Boston Dynamics, where they were involved in pioneering projects like large embodied multimodal models (PaLM-E, RT-2), scaling ChatGPT, and developing advanced robots like Atlas and Spot. Andy Zeng, a multiple-award-winning researcher in robotics from Princeton and UC Berkeley, previously worked on training large-scale robotics foundation models at Google. This collective background in both large-scale AI and cutting-edge robotics forms the core of Generalist AI's approach. The company is focused on developing "embodied foundation models," with an initial emphasis on dexterity, to enable robots to intelligently interact with the physical world. This involves significant advancements in data, models, and hardware. The business operates in the industrial robotics and artificial intelligence sectors, developing a platform that provides robots with end-to-end visuomotor control and dexterous sensorimotor policies. Their approach is to create AI models that empower robots to autonomously perform a wide array of complex manipulation tasks requiring fine motor control, bi-manual coordination, and sequential reasoning. This positions them to serve clients across various industries and eventually in homes, where humans and machines will work together. The company is currently in stealth mode but has shared research previews of its work. A key component of Generalist AI's strategy is the collection and utilization of massive amounts of real-world physical interaction data, as opposed to relying solely on simulations. Their flagship model, GEN-0, is a new class of foundation model trained on an extensive dataset, which started with 270,000 hours of real-world robot manipulation. The company has built an industrial-scale data collection operation, adding thousands of new hours of robot experience weekly, with the infrastructure to process this "internet-scale" physical data. They have stated that their research indicates a "phase transition" in model capability when crossing a 7-billion parameter threshold, which unlocks genuine adaptability and physical intelligence. The company has also developed internal tools to visualize and search this vast "universe of manipulation" data. Generalist AI has secured significant financial backing, raising $12.5 million in a Seed round and $128 million in a Series A round in early 2025, bringing its total funding to over $140 million. Key investors include NVentures, Boldstart Ventures, and Nvidia, indicating strong market confidence in their approach. Keywords: Generalist AI, general-purpose robots, embodied foundation models, robotics, artificial intelligence, dexterous manipulation, AI robotics, sensorimotor policies, visuomotor control, Andy Zeng, Pete Florence, Andrew Barry, GEN-0, physical interaction data, large-scale AI, foundation models for robotics, machine learning, robot learning, autonomous robots, real-world data training, robotics hardware, AI models, Series A, venture capital, Nvidia, Boldstart Ventures