Nebius and Positronic launch the Physical AI Competition: from simulation to a real robot

Nebius and Positronic Robotics have launched the Physical AI Competition, inviting teams to develop robotic manipulation policies in simulation by 13 December, after which the top 10 submissions will be evaluated on a real robot arm for more than $20,000 in Nebius GPU and Positronic evaluation credits. Using the widely adopted DROID setup—a Franka arm with a Robotiq gripper—and public training data, the competition applies identical tasks and scoring rules to all entrants, with independent real-world evaluation conducted by Positronic through PhAIL.ai. The initiative addresses the lack of comparable robotics benchmarks, as current open models range from roughly 12% to 70% average success on more than 100 real-robot everyday tasks. Teams can train on Nebius AI Cloud, submit models as containers, and access real-hardware testing without owning a robot lab, while early registered participants may receive GPU credits.

No tracked trend matches for this post yet.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.