AI News Daily한국어
Saturday, November 14, 2026

Story 3 of 12

OpenAI folds its standalone robotics team as general model Astra hits 74.8% on a robot benchmark versus 51% for specialist models

First seen on X 19 hours ago@deredleritt3r ♥ 9535 views

OpenAI folds its standalone robotics team as general model Astra hits 74.8% on a robot benchmark versus 51% for specialist modelsOpenAI folds its standalone robotics team as general model Astra hits 74.8% on a robot benchmark versus 51% for specialist modelsOpenAI folds its standalone robotics team as general model Astra hits 74.8% on a robot benchmark versus 51% for specialist modelsOpenAI folds its standalone robotics team as general model Astra hits 74.8% on a robot benchmark versus 51% for specialist models

OpenAI has disbanded its standalone robotics team and moved the researchers elsewhere, betting that its general-purpose models can now run robots directly.

According to a Fortune exclusive on Nov. 13, the robotics research group now sits in the main research team under Chief Research Officer Mark Chen, while applied robotics and hardware moved into the Applied organization under president Greg Brockman. A spokesperson said robotics remains core to OpenAI's long-term vision and called the change a sign of commitment, not retreat. The move was so recent that OpenAI still had booth #208 reserved at the CoRL robotics conference in Austin this week; when Fortune visited, it sat empty.

OpenAI said that early this year it expected to build robotics on top of Sora, but frontier models such as GPT-6 Astra proved directly usable on robots. In an Nvidia study, an Astra-powered agent hit a 74.8% success rate on a simulated manipulation benchmark, versus 51.0% for the best specialized robot models. OpenAI also shut a robotics team in 2020 and rebuilt one in 2024; its Figure partnership ended in February 2025. Sam Altman is targeting a first public demo of an in-house humanoid in 2027.

The X post relaying the news drew about 980 likes. Andrew Curran noted that "Anthropic reported a similar sudden and unusual jump in robotics capabilities with Fable 5.2." Others pushed back, sharing a post calling the headline "misleading" and saying the team is making big progress by leaning on Astra and newer models. MIT professor Russ Tedrake told Fortune that robot companies still hold an edge in real-world deployment data.

“Fortune is reporting that OpenAI is so convinced that LLMs will become brains for robots that it has disbanded its robotics team.”

Next story →Scientists hand Claude their writing, not their ideas: hypothesis work is 4.2% of 200,000 research sessionsRead an AI Insight →A CTO now spends more time promoting courses and MCing events than coding: Korean dev YouTuber Hyangro's 'third way' for junior developers

Get tomorrow's stories as an alert

We'll ping you when each new edition or breaking story lands. Free, no sign-up.

Daily morning AI news, new AI Insights, and a monthly briefing of top stories on the 1st.

More stories from the same day

  1. AI agents can now pay without seeing your card. The same week, United and Amazon are shutting agents out
  2. "Is this a trap?" Claude Fable 5.2 grew suspicious in a vending-machine test and earned less than GPT and Gemini
  3. Scientists hand Claude their writing, not their ideas: hypothesis work is 4.2% of 200,000 research sessions
  4. Russian prank duo posing as Ukraine's spy chief fooled three OpenAI security officials into a 17-minute call about offensive AI
  5. "20% of my 5-hour limit on one PR review": Claude Max and Ultra users say weekly limits have been cut in half
  6. "AI shouldn't be controlled by two companies": local AI engine llama.cpp hits 1.0 after three and a half years
  7. Nvidia shareholders Altimeter and Atreides urge board not to rush Jensen Huang's replacement, keep his return on the table
  8. Trump bought up to $5M in Nvidia stock days before scrapping China AI-chip curbs — and three more times after green-lighting H200 sales
  9. "Make the app cream-colored" is all it takes: Claude Code desktop themes hit 14,800 likes, and T3 Code ships the same idea hours later
  10. Novice hackers reach admin access in under 3 hours: FBI issues flash alert on open-source AI attack tool ARTEX, already used against Korean banks
  11. He used AI to find a 15-year-old bug and got rejected for it — while a Linux kernel maintainer steps back because half his review queue is AI patches
See all of Saturday, November 14, 2026 →

* Summaries and publishing on this page are done by an automated agent. Check the facts against the original links.

Privacy PolicySupportSend feedbackDiscord botClaude Code & CodexRSSListed on Sell With boost