Skip to main content

OpenAI disbands standalone robotics team and moves staff to other units

15 NOVEMBER 2026·2 MIN READ·2 SOURCES

OpenAI has disbanded its standalone robotics team, moving its researchers into the main research team and its applied robotics and hardware staff into the Applied organization. The company says the change is intended to accelerate robotics work, not abandon it.

Key takeaways · 4

  • 01

    Robotics researchers moved to OpenAI’s main research team under Chief Research Officer Mark Chen; applied robotics and hardware moved to Applied, which reports to President Greg Brockman.

  • 02

    OpenAI says general-purpose frontier models, including GPT-6 Astra, can be used directly for robotics.

  • 03

    An Astra-powered agent scored 74.8% on a simulated manipulation benchmark; that result is from a simulated benchmark, not evidence here of comparable physical-robot performance.

  • 04

    OpenAI’s stated rationale includes a discovery that motivated the change; Wojciech Zaremba also cited limited robot data and a more efficient path toward AGI.

A change in team structure

OpenAI moved its robotics researchers into the company’s main research team, under Chief Research Officer Mark Chen.[1] Applied robotics and hardware teams moved into the larger Applied organization, which reports to President Greg Brockman.[1] OpenAI describes the reorganization as a way to accelerate progress in robotics rather than abandon the work, and says robotics remains part of its long-term vision.[1] For teams tracking the company’s robotics work, the shift is organizational: dedicated staff are now within broader research and Applied units.[1]

The model strategy behind the move

OpenAI says general-purpose frontier models, including GPT-6 Astra, can be applied directly to robotics.[1] The company says a discovery motivated its decision to disband the standalone team.[1] Earlier, researchers had expected robotics progress to build on work such as Sora, OpenAI’s video-generation model.[1] Wojciech Zaremba later pointed to limited robot data and a more efficient path toward AGI.[1] The evidence describes the company’s reasoning, but does not specify the discovery or establish that one general-purpose model can handle every robotics task.

What the performance evidence shows

An Astra-powered agent achieved a 74.8% success rate on a simulated manipulation benchmark.[1] In a separate report covering 200 trials, Astra completed seven of 100 tasks, while MolmoAct2 completed none.[2] These results offer evidence about performance in the reported test settings; they do not establish how the models perform on physical robots in workplaces.[1][2] That distinction matters when assessing whether the team change signals readiness for deployment: the available benchmark figures are not, by themselves, proof of reliable real-world operation.

Robots, data and deployment

Robots are gaining work experience in real places, including factory floors.[1] OpenAI has said it is developing its own humanoid robot, with CEO Sam Altman targeting a first public demonstration in 2027.[1] OpenAI also faces pressure to rein in operating losses, reportedly $10.9 billion in the most recent quarter.[1]

For professionals evaluating robotics plans, the reorganization is a reason to test whether general-purpose models can reduce the need for dedicated robotics development—not a reason to assume that they already can. Keep simulated results separate from physical deployment evidence, and assess what data, integration work and task-specific validation your own use case requires.

Why it matters
For teams

AI training for your team

kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.

How this developed

  1. 15 November 2026

    OpenAI’s explanation now includes a discovery it says motivated disbanding the standalone team, its earlier expectation that robotics would build on work such as Sora, and Wojciech Zaremba’s later emphasis on limited robot data and a more efficient path to AGI; the evidence also adds context on real-world robot experience and reported operating-loss pressure.

  2. 15 November 2026

    OpenAI folds robotics work into research and Applied

Sources

AI fluency, one session a day, built for your work.