AW-0 debuts: one generative world action model plans in code, predicts, and controls robots
trevordarrell · x · 2026-10-27
AW-0 is billed as the first World Action Model that jointly generates code (explicit skill calls and control flow), future video, and continuous robot actions via joint denoising, letting semantic plans shape both predicted futures and motor control. It supports hybrid execution through perception/control APIs, adds only 40ms inference latency over the base WAM on an RTX 5090, and was pretrained on 3,000 hours of robot and human demonstrations. Paper and code are released.
Related event: UC Berkeley and NVIDIA Unveil AW-0, a Unified World Action Model(4 posts)→
More from Embodied
- DONIC hits 86.79% on bimanual dexterous tracking, crushing prior SOTA's 9.43% — liuziwei7 · 2026-10-27
- Bedrock Robotics raises $600M as its fully autonomous excavators do paid work on live job sites — Sethwinterroth · 2026-10-27
- AW-0 hits 80.8% on RoboTwin tasks, beating strongest baseline by 16 points — trevordarrell · 2026-10-27
- AW-0's code stream fixes world model prediction: picks mango where baseline grabs apple — trevordarrell · 2026-10-27
- AW-0 runs at 230ms on a single RTX 5090, 2.4x faster than autoregressive code planning — trevordarrell · 2026-10-27
- Skyfall AI's SKY-001 world model turns sketches into CAD builds in 3.9s, vs 199s for GPT-6 Astra — rohanpaul_ai · 2026-10-27