CALVIN ABC→D leaderboard

Long-horizon, language-conditioned benchmark that trains on environments A, B and C and tests zero-shot on unseen environment D. The official leaderboard is static, and recent papers report averages near 4.8 of 5.

Simulation · Franka Panda · PyBulletReport an error
4.78 / 5Best · MMaDA-VLA
24Models ranked
Near saturationSaturation
Nov 18, 2026Numbers checked
#ModelTypeScoreDateSource
1MMaDA-VLAWestlake UniversityVLA4.78 / 5, self-reported by the authors, not a leaderboard entryMar 2026
2Xiaomi-Robotics-0Xiaomi RoboticsVLA4.75 / 5, self-reported by the authors, not a leaderboard entryFeb 2026
3VITANanjing UniversityVLA4.73 / 5, second-hand number, not confirmed at the sourceNov 2025
4NS-VLABUPTVLA4.72 / 5, self-reported by the authors, not a leaderboard entryMar 2026
5π0.5Physical IntelligenceVLA4.69 / 5, self-reported by the authors, not a leaderboard entryJul 2026
6FLOWERKITVLA4.53 / 5, self-reported by the authors, not a leaderboard entrySep 2025
7Being-H0.7BeingBeyondVLA4.48 / 5, self-reported by the authors, not a leaderboard entryMay 2026
8X-VLATsinghua AIR et al.VLA4.43 / 5, self-reported by the authors, not a leaderboard entryOct 2025
9VLA-AdapterOpenHelix TeamVLA4.42 / 5, self-reported by the authors, not a leaderboard entrySep 2025
10UniVLA (BAAI)CASIA & BAAIVLA4.41 / 5, self-reported by the authors, not a leaderboard entryJun 2025
11VPP (Video Prediction Policy)Tsinghua University & RobotEraWorld-action4.33 / 5, self-reported by the authors, not a leaderboard entryDec 2024
12SeerShanghai AI LabPolicy4.28 / 5, self-reported by the authors, not a leaderboard entryDec 2024
13RoboVLMs (KosMos P.H.)Tsinghua University & ByteDance ResearchVLA4.25 / 5, self-reported by the authors, not a leaderboard entryDec 2024
14UP-VLATsinghua UniversityVLA4.08 / 5, self-reported by the authors, not a leaderboard entryJan 2025
15GR-MGByteDance ResearchPolicy4.04 / 5, self-reported by the authors, not a leaderboard entryAug 2024
16MoDEKITVLA4.01 / 5, self-reported by the authors, not a leaderboard entryDec 2024
17UniVLA (OpenDriveLab)OpenDriveLabVLA3.80 / 5, self-reported by the authors, not a leaderboard entryMay 2025
18RoboDualShanghai AI LabVLA3.66 / 5, self-reported by the authors, not a leaderboard entryOct 2024
19DitaShanghai AI LabPolicy3.61 / 5, self-reported by the authors, not a leaderboard entryMar 2025
203D Diffuser ActorCarnegie Mellon UniversityPolicy3.35 / 5, self-reported by the authors, not a leaderboard entryFeb 2024
21GR-1ByteDance ResearchPolicy3.06 / 5, self-reported by the authors, not a leaderboard entryDec 2023
22EnerVerseAGIBOT et al.World-action3.00 / 5, self-reported by the authors, not a leaderboard entryNov 2025
23Any3D-VLAGalbotVLA2.68 / 5, self-reported by the authors, not a leaderboard entryJan 2026
24RoboFlamingoByteDance ResearchVLA2.48 / 5, self-reported by the authors, not a leaderboard entryNov 2023

Metric: Average completed chain length (0–5) over 1,000 five-step instruction chains, ABC→D split. One result per model: RoboFYI re-run, then official leaderboard, then paper, then third party. Runs that deviate from the official protocol are listed on the model page but not ranked.

Wrong or missing number? Report an error