Primotion, a Shenzhen startup founded in July 2026 to build chips for embodied AI, has made its first public appearance not with silicon but with a model. Its embodied foundation model Paimon-0 ranks first on several public robot benchmarks, according to a detailed report by QbitAI published on Tencent News and an earlier report by The Beijing News.
Primotion says Paimon-0 scored 0.7754 to lead the RoboColiseum leaderboard, where it also topped the general manipulation and disturbance-adaptation sub-rankings, scored 58.1 overall on RoboCasa365 and posted a 0.73 average success rate on VLA-Arena. It also won a dual-arm robot policy challenge with 46.599 points. All of these results are self-reported by the company.
The model is built around a world action model, which predicts future states that an inverse dynamics model turns into actions. Rendering full future video is too slow for real-time robots, so Paimon-0 uses query tokens to distill only the core features, such as object 3D pose, semantics and affordance, before predicting actions on the device. Primotion says this cuts inference for a single action chunk from about 3,000 ms to 300 ms. The company stresses that this is chunk inference time, not the low-level control loop period. Pretraining relies heavily on first-person human manipulation video supplied by partner EgoScale.
Why would a chip company start with a model? CEO Ren Qingyuan, a former vice president at a RISC-V processor company, 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. Primotion splits work between a cloud planner it calls the meta brain and an on-device local brain that handles high-frequency feedback from vision, touch and force.
Hardware work has already begun. By adding an FPGA alongside a robot's main controller to offload pre- and post-processing, Primotion says it cut worst-case timing error on a 1 kHz EtherCAT control cycle from 30 microseconds in software mode to 152 nanoseconds. Much of that logic is meant to move into its own system-on-chip, which it plans to start delivering in 2028.
The company says it will not build robot bodies. It plans to open-source its progress model, Paimon-Eval, in November or December, and expects to release a second foundation model trained on first-person data in about three months.
SEE ALSO: Amap Open-Sources Lambda-0 · ShengShu Livestreams Motubrain 2.0 Training