← Leaderboard

Motus

World-action
Tsinghua University · Dec 15, 2025

Latent-action world model from Tsinghua with PKU and Horizon Robotics that unifies video generation and action prediction on a Wan2.2 backbone. Real-robot results are self-reported (63.22% average partial success on AC-One vs 14.79% for π0.5); it also reports RoboTwin 2.0.

Type
World-action
Parameters
7.1B
Backbone
Wan2.2-5B video model + Qwen3-VL-2B
Weights
Open
License
Apache-2.0
Commercial use
Allowed
Released
Dec 15, 2025
Robots
AC-One, AgileX ALOHA-2, Aloha-AgileX (sim)
SO-101
–
LeRobot
–
Inputs
–
VRAM
–

Benchmark results

1 of 13 benchmarks
RoboTwin 2.0Near saturationSimulation
  • 87.8%, self-reported by the authors, not a leaderboard entryPaper / model card ↗Dec 2025· self-reported, 50 clean + 500 randomized demos· n = 10000· non-standard
87.8%, self-reported by the authors, not a leaderboard entrynot ranked
Not reported on 12 benchmarks
Wrong or missing number? Report an error