Isaac 0.5 Released: 36B Dynamic MoE Open-Weight Embodied Foundation Model
code_star · x · 2026-08-27
Perceptron Inc has released Isaac 0.5, a 36B parameter dynamic MoE (Mixture of Experts) open-weight embodied foundation model. It integrates multimodal video understanding, embodied reasoning, and robot control into a single sparse backbone. Key highlights from the technical report include the null-expert MoE, joint VLM data mixtures, and scaling laws for transfer learning from unlabeled video data. Weights are available now on Hugging Face.
Related event: Perceptron Releases Isaac 0.5, a 36B Open-Weight Embodied Foundation Model(5 posts)→
More from Embodied
- Actuate 2026 recap: Data as infrastructure, real-world deployment as the test — Sethwinterroth · 2026-08-27
- NSF invests $90M in three new Science and Technology Centers, including human-robot co-adaptation — yudapearl · 2026-08-27
- Sai-Fi: Connects Meta Glasses to Autonomous Computer Agent — xwang_lk · 2026-08-27
- T800 robots battle it out at SF Robot Fight Night — verdakorz · 2026-08-27
- ECCV Paper: Spatially-Sparse Linear Attention for Low-Latency Event Detection — RexDouglass · 2026-08-27
- Figure's Index: 43k weekly users feeding robot data, $15M paid, $1B committed — coreylynch · 2026-08-27