OpenBMB Open-Sources MiniCPM-Robot: Embodied AI That Keeps Tracking Offline
iamfakhrealam · x · 2026-07-24
OpenBMB has open-sourced its first embodied AI model series, MiniCPM-Robot, at WAIC. The release features a wild live demo where unplugging a robot dog's network cable doesn't interrupt its tracking at all.
The series includes three main components:
- MiniCPM-RobotTrack: A 0.9B parameter model built on MiniCPM4-0.5B. It runs entirely on the robot's onboard hardware (e.g., Unitree Go2 Edu) with no internet required, achieving 180ms end-to-end latency at 5+ Hz.
- MiniCPM-RobotManip: A 1.5B general-purpose VLA model. Using efficient visual token compression, it matches the performance of Qwen-VLA and π0.5 while using roughly half the visual-processing compute.
- Inference Optimization: With PhyAI’s CUDA Graph optimization and custom Triton fused kernels, inference throughput jumps from 10 Hz to 33 Hz, and up to 36 Hz on NVIDIA H20.
Related event: OpenBMB Open-Sources MiniCPM-Robot for Offline Embodied AI(10 posts)→
More from Embodied
- Johns Hopkins Launches Full-Stack Hands-on Robot Learning Class with SO-101 Arm Kits — _krishna_murthy · 2026-09-11
- SyncWorld: In-Context Robot World Model Simulates Unseen Views and Embodiments Zero-Shot — ChongZzZhang · 2026-09-11
- A 3D Pose Dataset for Dogs Released — ducha_aiki · 2026-09-11
- Swaayatt demos autonomous driving at 52 km/h on mountain roads, self-recovers after skid — sanjeevs_iitr · 2026-09-11
- AUAR's MicroFactory brings a deployable robotic wood-panel factory to the construction site — lukas_m_ziegler · 2026-09-11
- Musk: Cybercab certified at 165 Wh/mi, the most efficient production EV ever — elonmusk · 2026-09-11