Wednesday, November 18, 2026

News

OpenAI disbands stand-alone robotics team, moves researchers into core divisions

ModelsPatryk Raba
OpenAI disbands stand-alone robotics team, moves researchers into core divisions
Fot. TechCrunch, Wikimedia Commons (CC BY 2.0)

OpenAI confirmed to Fortune that it has dissolved its stand-alone robotics team and split its staff between research and Applied. The company calls it an acceleration, not a retreat, arguing that general models such as GPT-6 Astra have proved directly useful in robotics.

Contents
  1. What changed
  2. The company's rationale
  3. History and financial pressure
  4. Skeptical voices

OpenAI has disbanded its stand-alone robotics team and moved its members to other parts of the company. It confirmed the change in response to questions from Fortune, which reported the story on November 13. According to a company spokesperson, the decision is meant to show commitment to robotics, not a retreat from it.

What changed

The robotics research team is now part of the main research organization, led by Chief Research Officer Mark Chen. The applied robotics and hardware teams have moved to the Applied organization, which reports to company President Greg Brockman.

The change was recent enough that OpenAI had already registered booth #208 at CoRL, a robotics conference in Austin. Fortune found it empty and without a logo. A few former members of the team were milling around the booth and still drew a crowd, despite the lack of a display or demo.

The company's rationale

OpenAI says that at the start of the year its researchers believed the best path was to build on work such as Sora, a video model that implicitly understands physics. Since then, the company says, general frontier models, including GPT-6 Astra, have proved directly useful in robotics, which prompted the change of approach.

As evidence, Fortune cites an Nvidia study. An agent built on Astra was given camera images, information about the robot and a control interface. On a simulated manipulation benchmark it achieved a 74.8% success rate, compared with 51.0% for the best specialized robotics models on that leaderboard.

Guanya Shi, an assistant professor at Carnegie Mellon's Robotics Institute, reads OpenAI's move as a signal that the company no longer treats robotics as a special problem. Shi believes OpenAI sees it as an application of general model intelligence instead.

History and financial pressure

This is not the first such move. OpenAI had an early robotics team that it shut down in 2020, and cofounder Wojciech Zaremba later pointed to limited robot data and a more efficient path to AGI through other methods; the company then focused on training large language models. The team was rebuilt in 2024, and the partnership with Figure on models for humanoids ended in February 2025, when Figure decided to build its own solution.

Fortune also ties the change to costs. Altman reportedly told employees last year to abandon side projects, and the company dropped some efforts, including Sora, to redirect computing power. According to the report, the operating loss in the most recent quarter reportedly reached $10.9 billion, and OpenAI is seeking a new funding round and preparing for a possible IPO in early 2027.

Skeptical voices

Russ Tedrake of MIT, cofounder and CEO of Walden Robotics, points out that robotics companies have an advantage because their robots already work in real places, such as factories. Deployment data helps adapt models to specific sites, and better foundation models make that data more valuable.

Fortune notes that large models are slow and run in the cloud, so they require a constant connection. Shi still sees a need for smaller specialized models that run on the robot's onboard computer. Pete Florence, cofounder of Generalist, says there may be many starting points for growing intelligence, and it is unclear which model will come to dominate robotics.

OpenAI says it is developing its own humanoid robot, and Sam Altman is targeting a first public demo in 2027. The debate over whether robots need models of their own touches startups that have raised billions on exactly that premise.

Share: