← Leaderboard

DYNA-2

World-action
Dyna Robotics · Aug 15, 2026

World-action model pretrained on over one million hours of egocentric human video that denoises future video and actions jointly. Results are self-reported: it passed 87% of production acceptance tests at unseen customer sites against 46% for DYNA-1, and its mean normalized score on 14 tasks rose from 20% to 53% as pretraining grew from 1k to 1M hours.

Type
World-action
Parameters
—
Backbone
Video diffusion mixture-of-transformers (DiT)
Weights
Closed
License
–
Commercial use
Unclear
Released
Aug 15, 2026
Robots
YAM (bimanual), YAM + WUJI-2 hands, semi-humanoid prototype
SO-101
–
LeRobot
–
Inputs
–
VRAM
–

Benchmark results

0 of 13 benchmarks

No benchmark results recorded yet.

Not reported on 13 benchmarks
Wrong or missing number? Report an error

Built on DYNA-2