Leaderboard
World ModelsProprietary
Veo 3
Google DeepMind's text/image-to-video model with native audio and real-world physics; among the top performers on the PAI-Bench physical-AI generation benchmark.
Crosshair Index
—
#6 of 10 · World Models
- Provider
- Google DeepMind
- Released
- 2025-05-20
- Parameters
- Undisclosed
Capability web
Strengths across world-modeling capabilities — each axis is one benchmark domain. Coverage is sparse while these evaluations mature; empty axes read “no data yet.”
Physical-AI Generation
82
skill
Generation quality plus physical plausibility for embodied / physical-AI scenes — driving, robotics, ego-centric — judged on PAI-Bench.
Scorecard
| Benchmark | Score | Source | Status |
|---|---|---|---|
| Something-Something v2 Motion Understanding | — | not evaluated | |
| EPIC-Kitchens-100 Anticipation Action Anticipation | — | not evaluated | |
| Perception Test Video Understanding | — | not evaluated | |
| Physics-IQ Generative Physics | — | not evaluated | |
| PAI-Bench-G Physical-AI Generation | 82.2 | PAI-Bench (arXiv:2512.01989, Table 3)paper | unverified |