OpenAI disbands robotics team as its general models start running robots
AI寒武纪 · wechat · 2026-11-14
OpenAI confirmed it has disbanded its standalone robotics team, folding researchers into the main research org under chief research officer Mark Chen, with the applied robotics/hardware team moving under president Greg Brockman. The trigger: general frontier models (reported as GPT-6 Astra) proved capable of directly controlling robots — an NVIDIA-simulated benchmark showed an Astra-driven agent hitting 74.8% success vs 51.0% for the best specialized robotics model, with no robot-specific training.
OpenAI denies retreating, saying robotics remains core, with Altman reportedly targeting a 2027 demo of its own humanoid. Experts are split: CMU's Guanya Shi sees robots becoming a downstream app of general intelligence (though onboard small models remain necessary for latency/offline), while MIT's Russ Tedrake argues real-world deployment data still gives robotics startups an edge. Context: OpenAI faces cost pressure with $10.9B quarterly operating losses ahead of a possible IPO, and the move challenges startups that raised billions on the premise robots need dedicated 'brains'.
Related event: OpenAI Reportedly Dissolves Robotics Team Months After Relaunch(20 posts)→
More from Companies & People
- Apple's device ecosystem may not be enough against better AI experiences — mobileraj · 2026-11-15
- A Third Path for Scientists: Boosting AI's Scientific Capabilities at Frontier Labs — kenbwork · 2026-11-15
- JP Morgan private bank publishes note on Nvidia's Jensen Huang — BenBajarin · 2026-11-15
- Free online textbook Principles of Operating Systems by Carl Burch covers interrupts to syscalls — burkov · 2026-11-15
- Delaware now only needs majority shareholder consent to convert a for-profit into a PBC — NathanpmYoung · 2026-11-15
- Google shipped three Gemini 4 variants since October; OpenAI released only gpt-6.1 terra — haider1 · 2026-11-15