🔍 Read the full analysis: OpenAI Splits Up Robotics: How Does Runway's Praxis-1 Fit In? on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
OpenAI has disbanded its stand-alone robotics team, moving its researchers and applied robotics staff into other company groups. The company says general-purpose models can now be applied directly to robotics; Runway’s open-weight Praxis-1 offers a faster, specialized alternative, though published tests do not establish which approach works better in real-world conditions.
OpenAI has broken up its stand-alone robotics team and moved its researchers and applied robotics staff into other parts of the company, according to Fortune, which said OpenAI confirmed the change. The reorganization comes as Runway releases Praxis-1, an open-weight model designed for robot control, renewing a key question for the industry: whether general-purpose AI can control robots effectively, or whether they need specialized models.
Fortune reported that OpenAI’s robotics researchers now sit within its main research organization under Chief Research Officer Mark Chen. Its applied robotics and hardware teams moved to the Applied organization, which reports to President Greg Brockman. The change was recent enough that OpenAI’s booth at the CoRL robotics conference in Austin remained on the floor plan, though Fortune found it empty. OpenAI has not published its own announcement about the reorganization.
An OpenAI spokesperson told Fortune the move reflects a change in how the company approaches robotics, not a retreat from it. The spokesperson said that general-purpose frontier models, including GPT-6 Astra, have shown they can be applied to robotics directly. OpenAI’s researchers had earlier expected to build on video models such as Sora. The company says robotics remains part of its long-term vision; the explanation and commitment were conveyed to Fortune, rather than in a public statement.
Runway released Praxis-1, a 27-billion-parameter “world action model”, two days before Fortune’s November 13 report. It is distributed under the Apache 2.0 license, with weights and code for fine-tuning. In the RoboLab-120 results cited by the reports, Praxis-1 scored 46.2%, the highest score among world action models and open-source robot models on the leaderboard, and fifth overall. The published comparison is based on benchmark testing, not a general finding that Praxis-1 is faster or more reliable in every robot setting.
Robot control loop · explainer
Who decides the robot’s next move?
Every robot repeats one loop: cameras see, a model decides, the arm moves. The two bets in this story put a different kind of model into the decide step.
Option 1
A general model in the loop
Example: GPT-6 Astra in NVIDIA’s simulation study
18.56 s per model call on average in the direct set-up. Inference was 97.0% of the estimated episode time; the simulation paused while the model was thinking.
Option 2
A specialist robot model in the loop
Example: Runway’s Praxis-1, a world action model with open weights
About 0.55–0.6 s per call on one H100 server, and each call returns 1.6 s of motion. Needs fine-tuning data for each robot.
General model plans the next stepslow, outside the loop
Option 3 · illustrative
Both, in two layers
The idea Runway and the NVIDIA researchers both describe
Runway calls the two approaches “complementary”: an agent plans, a fast policy carries out each step. NVIDIA’s authors say the best way to combine them “remains an open question”.
Sources: NVIDIA Seattle Robotics Lab, “Astranomy” study of GPT-6 Astra, Oct 14, 2026 (simulation, direct set-up); Runway, Praxis-1 technical report, Nov 2026 (one H100 server). Loops and timelines are schematic; the two figures come from different set-ups, not from a head-to-head test.
Graphic: Powered by Thorsten Meyer AI
Speed Shapes the Model Choice
The results highlight a practical trade-off between task success and response time. In a study by NVIDIA researchers, an agent using GPT-6 Astra achieved a 74.8% success rate on the simulated RoboLab-120 benchmark, compared with 51.0% for the best trained robot policy on that leaderboard. But in the study’s direct setup, each model call took an average of 18.56 seconds, and the simulation paused while the model generated its response. That delay matters for physical tasks where a robot may need to react continuously.
Runway says Praxis-1 takes about 0.55 to 0.6 seconds per call to produce 1.6 seconds of motion. In a separate test on one of its own robots, Runway reported that GPT-6 Astra took 15.6 times longer than Praxis-1 to complete three tasks and cost 51 times more per completed task. Those figures are Runway’s own test results; the material does not establish that the test was independently replicated or that its findings generalize to other robots and tasks.
For companies building robots, the choice affects more than benchmark scores: latency, inference expense, training and the ability to adapt a model all matter. The published evidence does not settle the question as a contest with one winner. General models may handle a wider range of tasks in simulation, while specialized models are designed to produce actions at a pace suited to control. The two approaches could also be combined.
Products Worth Considering
Samsung Galaxy Tab A11+ Kids Edition, 128 GB Android Tablet, Royal Blue
LG UltraGear 32GX850A-B 32-inch 4K UHD OLED Gaming Monitor
Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation
As an affiliate, we earn on qualifying purchases.
OpenAI’s Robotics Reorganization
This is the second time OpenAI has closed a stand-alone robotics team. The company wound down its first team in 2020. Cofounder Wojciech Zaremba later cited limited robot data and the prospect of pursuing artificial general intelligence more efficiently through other methods. OpenAI rebuilt a robotics team in 2024 and partnered with humanoid-robot maker Figure. Figure said in February 2025 that it would develop its own robot AI, ending that partnership.
OpenAI has said it is developing a humanoid robot, and Chief Executive Sam Altman has targeted a public demonstration in 2027. The restructuring report does not say that the robot project has been canceled. Fortune also described broader pressure on OpenAI to control costs and focus on core products, but the reason the company gave for this specific change was that its general-purpose models could be applied directly to robotics.
The performance figures come from different kinds of evidence. The NVIDIA researchers’ October 14, 2026, report was labeled a CoRL 2026 late-breaking result and tested GPT-6 Astra on RoboLab-120, a simulation of 120 tabletop tasks for one robot arm. Praxis-1’s leaderboard result and Runway’s robot test are separate evidence. These results should not be treated as a single controlled comparison across identical conditions.
“The two approaches are complementary.”
— Runway
Real-World Performance Still Unproven
The reported benchmarks do not show how either approach will perform across different physical robots, environments or longer tasks. RoboLab-120 is a simulation, and its results are not proof that a model can safely or reliably control a robot in everyday conditions. The source material also does not specify whether Praxis-1 and GPT-6 Astra were tested under identical hardware, prompts, task conditions and cost accounting.
It remains unclear how OpenAI will organize day-to-day robotics development after the split, whether it will continue building or deploying its own humanoid hardware, and whether the company will publish a formal account of the reorganization. OpenAI’s stated commitment and Altman’s demonstration target are not confirmation that a public demonstration will happen on schedule. The full cost and latency comparison also needs independent testing beyond Runway’s reported three-task trial.
Evidence to Watch in 2027
The next useful milestones are independent, comparable tests of general-purpose and specialized models on physical robots, including measures of response time, task completion, cost and reliability. Further Praxis-1 releases or evaluations could clarify how its open weights perform when adapted to other machines. OpenAI may also provide more detail on how its robotics work fits within its research and Applied organizations.
For OpenAI’s hardware plans, the stated public milestone is Altman’s target for a humanoid demonstration in 2027. No further details about the date, venue or capabilities were provided in the source material. Until those results and plans are clearer, the reorganization signals a shift in OpenAI’s approach, not a settled answer about which kind of model will control robots best.
Key Questions
What changed in OpenAI’s robotics organization?
According to Fortune, OpenAI moved robotics researchers into its main research organization under Mark Chen and moved applied robotics and hardware teams into the Applied organization under Greg Brockman. OpenAI confirmed the change to Fortune but had not posted its own announcement as of November 14, 2026.
What is Runway’s Praxis-1?
Praxis-1 is a 27-billion-parameter robot model released with open weights and code under the Apache 2.0 license. Runway describes it as a “world action model” and says it can be fine-tuned for use.
Did GPT-6 Astra outperform Praxis-1?
Not in a direct, like-for-like comparison established by the cited results. GPT-6 Astra scored 74.8% on the simulated RoboLab-120 benchmark, while Praxis-1 scored 46.2% on the leaderboard. The reported figures involve different evaluations, and the source material does not establish identical test conditions.
Why might a specialized model still be useful?
The reported tests suggest a possible speed advantage. Praxis-1 was reported to take about 0.55 to 0.6 seconds per call, while GPT-6 Astra averaged 18.56 seconds per call in the NVIDIA study’s direct setup. Those measurements came from different tests and do not establish performance across all robots or tasks.
Has OpenAI canceled its humanoid robot plans?
The reported reorganization does not say that OpenAI canceled its robot plans. The company told Fortune it remains committed to general-purpose robots, and Sam Altman has set a 2027 target for a public demonstration. The timing and details of that demonstration remain unconfirmed beyond that target.
Source: ThorstenMeyerAI.com
