Robot Intelligence
Index

How it was built

Benchmarks. We included public robot leaderboards that publish a score for every model and were active in 2026. That covers benchmarks whose own teams run or verify each entry, such as RoboDojo, RoboChallenge, RoboArena, RoboLab, AgiBot's RoboColiseum, Robocurve's RoboDojo-RC and Artificial Analysis's AA-Robotics, and the benchmarks most often reported in recent model papers, such as LIBERO, SimplerEnv and CALVIN, which are self-reported. A benchmark feeds the ranking only when some of its models also appear on another benchmark. Four more sources were collected but not used, and the benchmark table explains why.

LLM agents. AA-Robotics tests only general LLM agents, which write code to drive a real arm with no robot training. RoboDojo-RC tests the same kind of agent alongside robot models. Both enter as real-robot benchmarks, and AA-Robotics carries less weight because it scores one attempt per task. RoboColiseum also tests LLM agents next to robot models, in simulation.

Who gets ranked. By default the overall ranking includes a model only if it has at least one real-robot result and one simulation result, which 27 models do. The coverage menu also shows any two benchmarks (49 models) or every model (68). The simulation ranking needs two simulated benchmarks. The real-robot ranking shows every model with a real result.

Ratings. Each benchmark is turned into head-to-head comparisons between the models it tested, and one rating per model is fitted to all of them on a chess-style scale where every model starts at 1500. The methodology walks through it step by step.

Task types

Beyond the overall rating, the index rates models on four types of task. We sorted each benchmark's own task and category scores into these types by hand, assigning each task to the type that best describes its main difficulty. Each type is then rated from those scores alone.

Benchmarks that report no breakdown count only toward the overall views. RoboArena counts toward pick and place because its evaluators mostly choose pick-and-place tasks.

Where benchmarks disagree

The same model often finishes near the top of one benchmark and near the bottom of another. Of the 18 models that finish top three somewhere, 9 finish in the bottom half elsewhere.

The two largest swings, DM0 and Spirit v1.5, involve models that were prepared very differently on two real-robot benchmarks. On RoboChallenge, their developers entered them as a separate model for each task (as declared on the board), with up to about 1,000 demonstrations per task available for training. On RoboDojo, the benchmark team fine-tuned both itself on 100 demonstrations per task. Dexmal's own single multi-task version of DM0 scores 49 on RoboChallenge, 9th of 22, still far above its RoboDojo finish. With two models, we cannot tell which of these differences matters most.

Several models that lead easy simulated benchmarks, such as OpenVLA-OFT and MolmoAct2 on LIBERO variants, fall near the bottom of harder ones. The benchmarks agree much more at the bottom of the field, where older models rank low everywhere.

Methodology

Each model gets one rating, built from head-to-head comparisons inside every benchmark it appears on. Every model starts at a score of 1500, and the gap between two ratings predicts how often one outscores the other.

1. Collect each benchmark's scores

We pulled every benchmark's public results and took one headline score per model from each, listed in the table at the end of this section. We matched the same model across benchmarks by hand, using name and version only. A model fine-tuned separately for two benchmarks still counts as one model.

2. Turn each benchmark into head-to-head matchups

Within a benchmark, every pair of models it tested is one matchup. Instead of a plain win or loss, each matchup records how likely it is that one model really scored higher, given how noisy that benchmark is.

Φ is the standard bell-curve probability, and noise is how much a score could move if the benchmark were run again. On RoboDojo's real robots, π0.5 scores 22.9 and X-VLA 7.6, each with a noise of 2.5 points. The gap is about four times the combined noise, so π0.5 counts as better with near certainty. GR00T N1.7 at 5.9 against π0 at 5.8 counts as 0.51, almost a tie.

3. Weight each benchmark

Each benchmark's weight is how independent its evaluator is multiplied by how repeatable its scores are. A model's matchups on one benchmark together count for about that weight, however many models the benchmark tested, so a 63-model board does not drown out a 10-model one. Benchmarks that report several separate tasks, such as MolmoSpaces, split their weight across those tasks.

4. Fit one rating per model

Two ratings predict a matchup's outcome the way chess ratings do.

We choose the set of ratings whose predictions best match all the weighted matchup outcomes at once, a standard method known as a Bradley–Terry model. Models that never met are placed through the models they share, so shared models act as bridges between benchmarks. A small pull toward 1500 keeps a model with little evidence from landing at an extreme. The overall, simulation, real-robot and four task-type rankings are each fitted separately, so a rating only compares models within one view.

5. Read a rating

A model rated 400 points above another is expected to outscore it on a benchmark they share about 9 times in 10. Ratings are relative. They show how models compare on the benchmarks that connect them, not an absolute skill level.

6. Measure uncertainty

We re-ran the whole fit 300 times. Each re-run drew the benchmarks at random with replacement, so some count twice and some drop out, and nudged every score by its noise. A model's rank range is where its rank falls in 90% of those re-runs. The leaderboard sorts by rating. Sorting by worst plausible rank instead uses the bottom of that range, so a model sits high only when its position holds up. A model tested on one benchmark gets a narrower range than it deserves, since no re-run can test it against a second.

7. Measure disagreement

For each benchmark in a view, we place the model by percentile among the models we matched there, from 100 for first to 0 for last. Spread is the standard deviation of those percentiles. DM0 finishes in the 90th percentile on RoboChallenge and near the bottom of both RoboDojo tracks, for a spread of 50. GPT-6 Astra finishes near the top of all five of its benchmarks, for a spread of 11. A model with one benchmark in the view has no spread.

8. Decide who gets a rank

Every model's comparisons inform everyone's ratings, but only some models receive a rank. By default the overall ranking requires at least one real-robot and one simulation result, the simulation ranking requires two simulated benchmarks, the real-robot ranking includes every model with a real result and the task types follow the overall rule.

Benchmark weights and noise

Weight is evaluator independence (1.0 when the benchmark's team runs or verifies every entry, 0.4 for paper-reported numbers) multiplied by repeatability (0.60 for AA-Robotics, with one attempt per task, and 0.70 for RoboChallenge, whose repeat runs correlate at 0.70, up to 0.95). RoboColiseum's evaluator weight is 0.8 because AgiBot runs it and also enters its own model, and its repeatability is 0.85 because it does not publish episodes per task. Giving every benchmark equal weight produces nearly the same order, with rank correlation 0.99 or higher in every view.

Setup

Acknowledgments

This index exists because benchmark teams build robot evaluations, run them with care and publish the results openly. Testing models on real robots is slow and expensive, and every number on this page started with their work. We are grateful to the teams behind each benchmark:

• The RoboDojo team, for the RoboDojo simulation and real-robot tracks, run by the benchmark team with blinded raters (paper)
• Robocurve, for RoboDojo-RC and its independent tests of LLM agents and robot models on the same tasks
• Artificial Analysis, for AA-Robotics
• Dexmal, for RoboChallenge and the repeat runs that let us measure its reliability (paper)
• The RoboArena team and the many labs that volunteer blind evaluations (leaderboard, paper)
• NVIDIA, for RoboLab-120 (paper)
• Ai2, for MolmoSpaces (paper) and for collecting paper-reported results in its VLA evaluation harness
• The RoboTwin 2.0 team (leaderboard, paper)
• The RoboCasa team, for RoboCasa365 and its verified submissions (paper)
• AgiBot, for RoboColiseum (paper) and the Genie Sim 3.0 platform it runs on
• The LIBERO-Pro2.0 team at HUST (benchmark, paper) and the LIBERO-Plus authors (benchmark, paper)
• MLL Lab at Northwestern and its collaborators, for Embodied First Exam
• The creators of LIBERO, SimplerEnv and CALVIN, and every model team that submits to these benchmarks or reports results in its papers

This is the first version of the index. We will add benchmarks as they publish results and correct errors as they are found.

We’d love to hear what you think! Send feedback, corrections or benchmark suggestions to contact@withplayground.ai.

Request an evaluation

Want your robot policy in the next version of the index? Tell us about it, and we’ll be in touch shortly.