OpenAI has disbanded its stand-alone robotics team and reassigned its staff to other parts of the company. The research team now reports to Chief Research Officer Mark Chen’s main research group. Applied robotics and hardware teams are moving to the larger Applied organization, which reports to President Greg Brockman.
OpenAI told Fortune the move reflects a change in how it expects to make progress, not a retreat from robotics. The company says its general-purpose models, including GPT-6 Astra, have shown they can be applied directly to robotics. It is also developing a humanoid robot, with CEO Sam Altman targeting a public demonstration in 2027.
Why is OpenAI changing course?
OpenAI previously had a robotics team in its early years, then wound it down in 2020. It rebuilt a stand-alone team in 2024 and partnered with humanoid-robot maker Figure to develop AI models. Figure said in February 2025 that it would develop its own robot AI, ending the partnership.
The new structure fits a broader industry bet: that models trained for tasks such as language, coding and video can also control machines, rather than requiring a separate AI model built just for robots. OpenAI told Fortune that researchers initially saw Sora, its video-generation model, as a route to robotics because creating realistic video requires some understanding of how the physical world works. The company says its newer general-purpose models changed that calculation.
What can general models do?
A Nvidia study reported a 74.8% success rate for an agent powered by GPT-6 Astra on a simulated robot-manipulation benchmark. The best specialized robotics models on that leaderboard were reported at 51.0%. The test used a simulated robot, not a machine acting in the real world, so it does not establish how well Astra performs in factories or homes.
Researchers say general models can adjust to changes such as a shifted bottle or different lighting, and reason through a sequence of actions. But they can be slow and depend on cloud computing, which means a robot may need a steady internet connection. Carnegie Mellon researcher Guanya Shi sees a continuing role for smaller models that run on a robot itself. MIT’s Russ Tedrake also points to an advantage for robotics companies: robots already working in places such as factories collect experience and data that can help them adapt to specific tasks.
Why it matters
For robot makers and factory operators, OpenAI’s move suggests general-purpose models may handle more tasks without a dedicated robotics model. But the reported Astra result was in simulation, and models that depend on cloud access may not suit every robot or workplace.
