OpenAI Disbands Robotics Team, Bets GPT-6 Astra Can Run Robots

OpenAI has dissolved its stand-alone robotics team for the second time in six years, reassigning researchers after its general-purpose models outperformed specialized robot brains on a benchmark. The company confirmed the reorganization to Fortune on November 13 and framed it as an acceleration, not a retreat.

The numbers explain the move. In an Nvidia study, researchers gave GPT-6 Astra camera images, robot state information, and a control interface to command a simulated robot. The Astra-powered agent achieved a 74.8% success rate on a simulated manipulation benchmark. The best specialized robotics models on that leaderboard reported 51.0%. OpenAI's conclusion: general frontier models can run robots directly, without a dedicated robotics stack.

Who now runs what?

The restructuring splits the old team in two. The robotics research group now reports to Chief Research Officer Mark Chen. Applied robotics and hardware teams move under President Greg Brockman, inside OpenAI's broader Applied division.

A spokesperson said robotics "remains a core part of OpenAI's long-term vision." The change, the company said, reflects a research discovery rather than a cost decision.

OpenAI's own thinking shifted during the year. At the start, researchers assumed the best path to robot control ran through video models such as Sora, which must hold an implicit understanding of physical dynamics to generate realistic footage. Since then, the spokesperson said, general-purpose frontier models showed they could handle robotics directly. That finding motivated the reorganization.

Why this is OpenAI's second robotics shutdown

This is the second disbandment in six years. OpenAI ran a robotics team in its early years, then wound it down in 2020. Cofounder Wojciech Zaremba later cited limited robot data and a more efficient path to artificial general intelligence through large language models — a pivot that produced ChatGPT in late 2022.

The company rebuilt a stand-alone team in 2024 and announced a partnership with humanoid startup Figure. That deal ended in February 2025, when Figure said it would develop its own robot AI. OpenAI has since said it is building its own humanoid, with CEO Sam Altman targeting a first public demonstration in 2027.

The timing showed at CoRL, a top robotics conference in Austin this week. OpenAI had registered booth #208. Fortune found it empty, with no company logo on display. A few former team members milled about and drew a crowd despite having no demo.

What does the bet mean for the industry?

OpenAI is wagering that models built for language, code, and vision can also drive "embodied AI" — robots, driverless cars, drones. On that view, separate teams building specialized robot brains become unnecessary.

Guanya Shi, an assistant professor at Carnegie Mellon's Robotics Institute, read the move as a signal that OpenAI sees robotics as "no longer a special problem." Instead, it becomes a downstream application of general intelligence, much as the same frontier models now power coding, legal, and financial tools. Shi said advances in spatial reasoning let general models cope with a bottle in a different position or changing lighting, and reason about which step went wrong.

Shi still sees a need for smaller, specialized models that run on a robot's onboard computer. Large language models are slow, typically cloud-hosted, and need a constant network connection.

Do robotics startups still have an edge?

Russ Tedrake, an MIT professor and CEO of Walden Robotics, argued yes. Robots at robotics companies are already accumulating work experience in real places such as factory floors. That deployment data helps robots adapt to specific workplaces, and better foundation models make site-specific data more valuable because robots learn from it faster.

Pete Florence, cofounder of Generalist, a startup training general-purpose AI models for robots, said it is unclear that any existing model will dominate robotics in the long run.

There are also caveats. The Nvidia result came from simulation, not a physical robot. An analysis by explainx.ai cautions that a single simulation benchmark does not cover contact-rich real-world tasks. OpenAI did not disclose how many researchers the reassignment affects.

What does the money context look like?

The reorganization lands amid pressure on OpenAI to rein in costs and focus on core products. Altman told staff last year to abandon "side quests," and the company has scaled back other efforts, including Sora, to reallocate computing resources. Fortune reported operating losses of $10.9 billion in the most recent quarter, as OpenAI seeks to close a new funding round and prepares for a possible IPO in early 2027.

OpenAI's statement emphasized research motivation, not cost pressure. For robotics startups that raised billions on the premise that robots need bespoke brains, the challenge is direct: if GPT-6 Astra keeps improving, the value shifts from building model architectures to owning real-world deployment data.