RoboArena leaderboard

Distributed real-robot arena in which evaluators at different sites run blind pairwise comparisons of DROID policies on tasks they choose. Ratings shift between snapshots and are not comparable across dates.

Real robot · DROID FrankaReport an error
1788Best · Spirit v1.6
13Models ranked
OpenSaturation
Nov 18, 2026Numbers checked
#ModelTypeRatingDateSource
1Spirit v1.6Spirit AIVLA1788Nov 17, 2026
2DreamZeroNVIDIAWorld-action1735Nov 17, 2026
3ApricotUnknownVLA1661Nov 17, 2026
4π0.5Physical IntelligenceVLA1608Nov 17, 2026
5π0-FASTPhysical IntelligenceVLA1582Nov 17, 2026
6Galaxea G0.5Galaxea AIVLA1559Nov 17, 2026
7PaliGemma (FSQ)Physical IntelligenceVLA1546Nov 17, 2026
8PaliGemma (flow)Physical IntelligenceVLA1533Nov 17, 2026
9PaliGemma (FAST, DROID tokenizer)Physical IntelligenceVLA1531Nov 17, 2026
10PaliGemma (FAST+)Physical IntelligenceVLA1516Nov 17, 2026
11damUnknownVLA1494Nov 17, 2026
12π0Physical IntelligenceVLA1461Nov 17, 2026
13PaliGemma (binning)Physical IntelligenceVLA786Nov 17, 2026

Metric: Bradley-Terry rating from double-blind pairwise A/B evaluations; policies with <20 evals hidden. 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