OpenAI Disbands Its Robotics Team Again, Betting GPT-6 Astra Can Run Robots Directly

OpenAI has disbanded its stand-alone robotics team, the company confirmed to Fortune on Friday, November 13, 2026, in the second such closure in six years. This time, though, OpenAI is framing the move as a vote of confidence rather than a retreat: its general-purpose frontier models, including GPT-6 Astra, can now be applied directly to robots, so a separate team building robot-specific models is no longer the fastest way forward.

The robotics researchers are moving into OpenAI's main research organisation, reporting to Chief Research Officer Mark Chen, while the applied robotics and hardware staff join the company's larger Applied organisation under President Greg Brockman. As a unit, the robotics team no longer exists. OpenAI has published no blog post about the change, and it has not said how many people were affected.

What OpenAI says changed

According to Fortune's exclusive report, the reorganisation reflects a shift in how OpenAI's researchers think about the problem. At the start of the year, the assumption inside the company was that robotics should build on video models such as Sora, which carry an implicit picture of how the physical world works. Since then, an OpenAI spokesperson told Fortune, general-purpose frontier models including GPT-6 Astra have shown they can be applied directly to robotics, and that finding motivated folding the researchers into the teams building more general models.

The spokesperson stressed that robotics remains core to OpenAI's long-term vision and that the company is still committed to bringing general-purpose robots into the world. Chief Executive Sam Altman is still targeting a first public demonstration of OpenAI's own humanoid robot in 2027, according to Fortune.

The timing was visible at this week's Conference on Robot Learning (CoRL) in Austin, Texas. OpenAI had already registered a booth, number 208, but Fortune found it empty, with no company logo, even as a few former team members still drew a crowd on the conference floor.

The evidence OpenAI is pointing to

The strongest data point in Fortune's report comes from an Nvidia study published on October 14, 2026. Researchers gave GPT-6 Astra camera images, robot state information and a control interface so the model could command a robot directly. On RoboLab-120, a simulated pick-and-place manipulation benchmark, the Astra-powered agent reached a 74.8 percent success rate, against 51.0 percent for the best dedicated robot model on the same leaderboard.

Two cautions apply. The benchmark was a simulation, not a physical robot, and a single benchmark does not cover contact-rich, real-world tasks. Follow-on reporting on November 14 added a hardware data point: OpenAI's own StationeryBench comparison, which pits GPT-6 Astra against Ai2's MolmoAct2 vision-language-action model on five bimanual desk tasks with real dual-arm robots, reported 7 full task completions out of 100 trials for Astra and none for MolmoAct2, with a mean progress score of 46 against 12. Those are directional results, not a solved manipulation story, but they point the same way as the Nvidia simulation.

A third act for OpenAI robotics

This is the third structure for robotics at OpenAI in six years. The company had a robotics team in its early days and wound it down in 2020, with co-founder Wojciech Zaremba later citing limited robot data and a more efficient path to AGI through other methods. OpenAI rebuilt a stand-alone team in 2024 and announced a partnership with Figure to build models for humanoid robots; that ended in February 2025, when Figure said it would develop its own robot AI. OpenAI has since said it is building its own humanoid.

Fortune also reported business context around the move: the company is under pressure to cut costs and focus on core products, with operating losses of $10.9 billion in the most recent quarter and a possible IPO in early 2027. OpenAI did not tie the robotics reorganisation to costs in the statement quoted by Fortune, and the spokesperson framed it as a way to accelerate progress.

Experts are split on the bet

Fortune quoted three robotics voices, and they do not fully agree. Guanya Shi, an assistant professor at Carnegie Mellon's Robotics Institute, reads the move as a signal that OpenAI sees robotics as no longer a special problem, noting that general models can already cope when a bottle moves or the lighting changes, and can reason about which step went wrong. He still sees a need for smaller specialised models that run on the robot without a constant network link, because large models are slow and cloud-hosted.

Russ Tedrake, an MIT professor and chief executive of Walden Robotics, argues that robotics companies keep an edge because their robots already work in real places such as factory floors, and that deployment data becomes more valuable as foundation models improve. Pete Florence, co-founder of Generalist, says it is unclear that any existing model will dominate robotics in the long run.

What it means for robotics builders

Many robotics startups have raised money on the premise that robots need their own brains. OpenAI's move challenges that assumption. If a general model plus a control interface works, specialised robot models face a harder case to make. The counter-arguments are real, though: deployment data from robots working in the field, and the latency and offline-operation limits of large cloud models, both favour keeping smaller models on the machine.

The same week also showed the field moving in both directions. Runway released Praxis-1, an open-weight dedicated robot model, two days before Fortune's report, while OpenAI's own work points the other way. For readers following the field, the signals to watch are the 2027 humanoid demonstration, the next benchmarks using general models on real hardware, and whether other labs make the same structural move.

The reorganisation also sits alongside OpenAI's wider push to put its models to work in everyday tasks. This week the company launched ChatGPT Wallet, letting AI agents shop and check out for users — another case of a general model being pointed at a job that once needed a specialist.

Sources

• Fortune, "OpenAI disbands its stand-alone robotics team as its general models start running robots", Wen Shao, November 13, 2026 (exclusive; via syndicated coverage)
• ExplainX, "OpenAI Disbands Robotics Team: GPT-6 Astra Bet (2026)", November 14, 2026
• Dev.to, "OpenAI Disbanded Its Robotics Team. GPT-6 Astra Is the Robot Now", November 14, 2026
• Singularity.kiwi, "OpenAI Folds Its Robotics Team Into The Mothership", November 14, 2026