Primotion, a Shenzhen startup founded in July 2026 to build chips for embodied AI, has made its first public appearance with an embodied foundation model that ranks first on several robot benchmarks.

• Primotion says Paimon-0 scored 0.7754 to lead the RoboColiseum leaderboard, where it also topped the general manipulation and disturbance-adaptation sub-rankings.
• The model uses query tokens to distill only core features such as object 3D pose, semantics and affordance before predicting actions on the device, cutting inference for a single action chunk from about 3,000 ms to 300 ms.
• CEO Ren Qingyuan argues that a team must first run the full loop of data, model, agent and evaluation on real customer tasks, then use those workloads to define the chip.