🔍 Read the full analysis: OpenAI Splits Its Robotics Team As Praxis-1 Brings Robot AI Into View on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
OpenAI has dissolved its stand-alone robotics team, moving researchers into its main research group and applied robotics and hardware staff into its Applied organization. The company says general-purpose models can now be used directly for robotics; Runway’s new open-weight Praxis-1, meanwhile, highlights a trade-off between broad task performance and faster robot control.
OpenAI has broken up its stand-alone robotics team, moving its researchers into the company’s main research group and its applied robotics and hardware teams into its Applied organization, Fortune reported on November 13. The reorganization came into view at the CoRL robotics conference in Austin, where the company’s listed booth was empty, as Runway introduced Praxis-1, an open-weight model built specifically for robot action.
OpenAI confirmed the team change to Fortune when the outlet asked about the empty booth. According to the report, research staff now report to Chief Research Officer Mark Chen, while applied robotics and hardware staff moved to the Applied organization under President Greg Brockman. OpenAI did not publish a separate announcement about the change, and its spokesperson described it as a commitment to robotics rather than a retreat.
The spokesperson told Fortune that OpenAI’s view had changed as its general-purpose frontier models, including GPT-6 Astra, showed they could be applied to robotics directly. Earlier in the year, researchers expected to build on video models such as Sora. The company says its broader models now offer a route to robot control without maintaining a separate stand-alone research team.
Runway released Praxis-1 two days before the Fortune report. The 27-billion-parameter “world action model” is available with weights and fine-tuning code under the Apache 2.0 license, which permits commercial use under that license. In results reported in the source material, Praxis-1 scored 46.2% on the RoboLab-120 benchmark, the highest score among world action models and open-source robot models listed there, and fifth overall.
A separate NVIDIA research study tested GPT-6 Astra on the same simulation benchmark. Its agentic setup recorded a 74.8% success rate, compared with 51.0% for the top trained robot policy cited on the leaderboard. That result came with a major speed limitation: in the study’s direct setup, a model call averaged 18.56 seconds, and the simulation paused while the model generated its response.
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
Robot AI Faces a Speed Trade-Off
The results frame the practical question raised by the two developments: whether robot builders should use general-purpose models, purpose-built systems, or a combination. In the reported simulation, GPT-6 Astra completed a larger share of tasks in its strongest setup, while Praxis-1’s reported calls took about 0.55 to 0.6 seconds to produce 1.6 seconds of motion. That difference matters for physical robots, which must respond to changing conditions rather than wait many seconds between decisions.
Runway also compared the models on one of its robots. It said 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 come from Runway’s own test and should not be treated as an independent, broad comparison. Still, the contrast points to a deployment issue: benchmark success alone does not show whether a model can operate quickly or affordably in a real robot.
For OpenAI, folding robotics staff into larger organizations suggests the company sees robot work as connected to development of its general models and applied products. The change does not establish that OpenAI has stopped building robots. The spokesperson said robotics remains part of its long-term vision, while the reported performance gap between systems suggests that specialist models may remain useful even as general models improve.
Products Worth Considering
Apple MacBook Air 15-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, Midnight
Apple MacBook Neo 13-inch Laptop with A18 Pro Chip, 8GB Memory, 512GB SSD, Liquid Retina Display, and Touch ID – Blush
FLEXISPOT Electric Adjustable Workbench, 48 x 30 in, Maple
As an affiliate, we earn on qualifying purchases.
OpenAI’s Robotics Team History
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 pointed to limited robot data and a more efficient path to artificial general intelligence through other methods. OpenAI rebuilt a robotics team in 2024 and partnered with humanoid maker Figure; Figure said in February 2025 that it would develop its own robot AI, ending that partnership.
The change follows OpenAI’s renewed work on humanoid robotics. Fortune reported that the company is developing its own humanoid robot and that CEO Sam Altman has targeted a public demonstration in 2027. That target is a stated plan, not confirmation that a demonstration will happen on schedule. Fortune also described pressure on OpenAI to control costs and focus on core products, but the explanation the company gave for this reorganization was that general-purpose models had become applicable to robotics.
The performance evidence comes from a study by researchers at NVIDIA’s Seattle Robotics Lab, published October 14 as a CoRL 2026 late-breaking result. It tested a single robot arm in RoboLab-120, a simulation of 120 tabletop tasks. The researchers evaluated GPT-6 Astra as a robot policy that used camera images, robot information and a control interface, without robot-specific training. These are simulation results, not proof of equivalent performance on physical robots.
“Complementary.”
— Runway, describing the relationship between general and specialist models
Deployment Results Still Unknown
The reported benchmark results do not establish how either model performs across different robot bodies, environments or unscripted tasks. RoboLab-120 is a simulation, and the source material does not provide independent, head-to-head testing of GPT-6 Astra and Praxis-1 across a shared set of real-world robots. Runway’s timing and cost comparison is its own test of one robot completing three tasks.
OpenAI has not published its own detailed account of the restructuring, and it is not clear how many staff members were moved, what projects will continue or whether the company plans to release a robot model or hardware product. The reported 2027 demonstration target has not been accompanied here by a confirmed date or further public details. It also remains unclear whether general-purpose models can reduce response times enough for reliable real-time control.
OpenAI’s 2027 Robot Milestone
The next public marker identified in the reporting is OpenAI’s stated aim to hold a first public demonstration of its humanoid robot in 2027. The company has not announced a date or described what the demonstration would show. Until then, further information about the reorganized teams, model capabilities and hardware plans would help clarify what work is continuing.
For Praxis-1 and the broader model comparison, the relevant next evidence will be results beyond the reported benchmark: independent testing, performance on physical robots and details about speed, reliability and cost under comparable conditions. Those results could show whether dedicated models are needed for fast control or whether general-purpose systems can perform the same tasks within practical operating limits.
Key Questions
What did OpenAI change in its robotics organization?
According to Fortune, OpenAI moved its robotics researchers into the main research organization under Mark Chen and its applied robotics and hardware teams into the Applied organization under Greg Brockman. OpenAI confirmed the change to Fortune when asked.
Has OpenAI abandoned robotics?
The company’s spokesperson said robotics remains part of OpenAI’s long-term vision and described the reorganization as a sign of commitment, not a retreat. OpenAI has not published a separate detailed statement about the change.
What is Praxis-1?
Praxis-1 is Runway’s 27-billion-parameter robot model, released with weights and fine-tuning code under the Apache 2.0 license. Runway describes it as a “world action model.”
Which model performed better on RoboLab-120?
In the reported NVIDIA simulation study, GPT-6 Astra’s agentic setup reached 74.8% success. Praxis-1 scored 46.2% in the results reported by the source material. These results come from different reported evaluations and do not establish a direct, independent head-to-head comparison.
When might OpenAI show its humanoid robot?
Fortune reported that CEO Sam Altman has targeted a public demonstration in 2027. No specific date or demonstration details were provided, so the timing remains a target rather than a confirmed event.
Source: ThorstenMeyerAI.com
