Embodied brain model trained on a 1,000-GPU cluster, says post
teortaxesTex · x · 2026-07-22
The post cites a technical update saying an embodied-intelligence brain model was trained on a domestic 1,000-card cluster.
It also claims that the full-stack native training of EvoPhys-World, described as the world’s first global 5D world model from Peking University’s EvoPhys team, was supported at the infrastructure layer by the MUSA software stack.
More from Embodied
- Sim2real and evaluation are becoming central to robotics simulation work — m_wulfmeier · 2026-07-22
- VideoRebirth says its video-native Olympus world model can power simulation agents — 机器之心 · 2026-07-22
- Snap Spectacles demo turns World Cup tracking data into editable photoreal replay — stspanho · 2026-07-22
- Sam Altman and Jony Ive’s hardware project becomes part of OpenAI’s internal fight — 2C_ornot2C · 2026-07-22
- Robot manipulation needs mobile navigation and standardized benchmarks, researchers argue — YuXiang_IRVL · 2026-07-22
- Hikvision pitches physical AI with multimodal sensing and edge-cloud deployment — 智东西 · 2026-07-22