← Leaderboard

UniVLA (BAAI)

VLA
CASIA & BAAI · Jun 24, 2025

“Unified Vision-Language-Action Model” (Wang et al.): an 8.5B autoregressive transformer on Emu3 that models vision, language and FAST action tokens in one sequence. It is a different model from OpenDriveLab’s UniVLA.

Type
VLA
Parameters
8.5B
Backbone
Emu3
Weights
Open
License
Apache-2.0
Commercial use
Unclear
Released
Jun 24, 2025
Robots
Franka Panda (sim), WidowX (Bridge), ALOHA
SO-101
–
LeRobot
–

Benchmark results

3 of 13 benchmarks
CALVIN ABC→DNear saturationSimulation
#10 of 114.41 / 5, self-reported by the authors, not a leaderboard entry
LIBEROSaturatedSimulation
#38 of 4695.5%, self-reported by the authors, not a leaderboard entry
SimplerEnv WidowXSaturatedSimulation
#12 of 1969.8%, self-reported by the authors, not a leaderboard entry
Not reported on 10 benchmarks

Adoption

updated Nov 17, 2026
HF downloads · 30d
–
HF likes
8
GitHub stars
324
GitHub forks
22
Wrong or missing number? Report an error