Orca World Model Learns Robotics From Video
The Decoder · rss · 2026-07-11
BAAIJ has released Orca, a world model designed for robotic tasks. Its key feature is predicting abstract world states rather than tokens or pixels.
Key Information
- Training data: 125,000 hours of video
- Training process: No action labels were used
- Results: Orca matched the performance of the specialized π0.5 system across 5 robotic tasks
Significance
This approach could alleviate the long-standing data scarcity problem in robotics. By learning the structural dynamics of the world directly from video, it bypasses the need for expensive action annotations.
More from Embodied
- NVIDIA open-sources FLORA, a robotics data engine ~125x cheaper than GPT-6 Astra — nvidia · 2026-10-27
- Embodied AI startups need a different operational discipline, investor says — ___Mufasaa · 2026-10-27
- π0.5 study: raising demos from 10 to 438 lifts Place Ball in Bowl success from 7% to 83% — stingning · 2026-10-27
- GPT-6 Astra controls dexterous hands zero-shot at just 5.2% success in 500 trials — stingning · 2026-10-27
- HandsOn benchmark: 66 verified tasks and 18,000 human demos for dexterous manipulation — stingning · 2026-10-27
- DexPort: a handheld interface with a real robot hand for dexterous-hand data collection — pabbeel · 2026-10-27