Isaac 0.5: open-source 36B embodied foundation model trained on 1M hours of video, weights released
AkshatS07 · x · 2026-09-03
Perceptron released Isaac 0.5, an open-source embodied foundation model it calls the first open model at the frontier of multimodal video understanding, embodied reasoning, and robot control. Key facts: a 36B-parameter sparse model ingesting images, video, language, robot state and past actions; trained on 100K hours of robot experience, 1M hours of general video, 3T multimodal tokens across 35+ embodiments. They propose a scaling law for video-vs-teleop data mixes: scaling general video from 1K to 1M hours cut teleoperation needed to hit the same held-out action loss from 5,900 hours to 28. Weights and paper are public; YAM and SO-101 checkpoints are coming.
More from Embodied
- Yacine says RNNs learn sysid anyway and vision policies aren't that hard — yacineMTB · 2026-09-03
- Robotics' demo-to-reality gap will dwarf LLMs', argues new post listing 14 challenges — Ben_Reinhardt · 2026-09-03
- Jetson Thor day 1: local VLM + RealSense 3D vision describes scenes every 350ms — chrismatthieu · 2026-09-03
- Video of an 80-year-old with a cane tailed by a robot goes viral — TansuYegen · 2026-09-03
- Viral robot attack video exposed as staged: a button on its shoulder triggered it — 2C_ornot2C · 2026-09-03
- Robot in Russian tech store tries to fight back after customer pushes it — _onionesque · 2026-09-03