RoboColosseum
Physical robots execute fresh instructions in paired blind trials. Human evaluators choose the better run or a tie, and comparisons determine ratings for each embodiment.
The separate fine-tune track uses 100 demonstrations per task and 25 attempts, reporting success and progress. Ratings cannot be compared across embodiments or tracks.
- Made by
- RoboColosseum consortium
- Released
- November 2026
- Kind
- Real robots
- Level
- Everyday
- Scored
- Ranked by votes
- Answers
- Agent task
- Tasks
- Public, plus a private test set
- Measures
- Preference
- Language
- English
- In and out
- Text, Images, Time series → Actions
- Interaction
- Agentic
- Environment
- A physical robot or device
- Items from
- Live play or votes
- Refresh
- Live
- Run by
- A host
- In views
- Agents, Vision, Robotics, Kept fresh
In the authors’ words
General-purpose robot policies must work across robots, tasks, and environments, yet the variation that makes physical trials informative also makes them difficult to reproduce. We introduce RoboColosseum, a distributed evaluation framework that combines unified policy serving and standardized protocols across a network of locally operated robots in diverse environments. Our initial study evaluates five vision-language-action (VLA) policies and three frontier AI agents across four embodiments (SO-101, Franka, Bimanual-YAM, and G1 Humanoid). The Open Track compares models without task-specific adaptation in blinded, pairwise trials under matched local conditions. The Fine-tune Track measures how well VLA policies adapt to 20 tasks across all four embodiments when each policy receives the same 100 demonstrations per task. Across 5,699 pairwise Open Track comparisons at 124 sites, frontier agents rival specialized VLA policies, while all systems struggle to complete open-world tasks reliably. After fine-tuning, GR00T N1.7 leads on SO-101 and pi0.5 leads on Franka and Bimanual-YAM. Policy rankings show weak agreement across embodiment-specific task suites (Kendall W = 0.24). Open Track rankings poorly predict relative performance after fine-tuning, with substantial rank reversals on all three embodiments shared between tracks. Measuring partial progress reveals differences that task success alone misses. On the G1 Humanoid, no policy completes any of the five tasks, yet average progress ranges from 13.3% to 23.2% across policies. Open Track rankings stabilize as trials accumulate, and active matchmaking achieves comparable ranking agreement with substantially fewer physical trials than uniform sampling. These results show how distributed physical evaluation can reveal where robot policies generalize, how well they adapt to new tasks, and how far they progress on tasks they fail to complete.
gpt-6-astra-droid tops the leaderboard with 1,925
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. The board ranks policies separately for each robot. The headline is the Franka Open track rating, which the board shows first. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.
Bars start at 1,000.
- 1 gpt-6-astra-droid rig-v1 1,925
- 2 molmoact2-droid d8c1abd8a27d8e859455bbe514df2bcc617db0fb 1,869
- 3 g05-droid e312be81e90c56a55bcb26b57429bd39a335b449 1,744
- 4 claude-opus-5-5-droid rig-v1 1,636
- 5 gr00t-n17-droid 05e7cc97e40dbd33b0890c35cc0214fcb0547ab5 1,350
- 6 pi05-droid 72824c0a93f00ce5bb8bedb7feb58953ba1da364 1,285
- 7 lingbot-vla-v2-6b-droid 11c703bf6a5c1f45b3b69168482da11fdbba53d7 1,233
- 8 grok-4.7-droid 4.7 1,160
Source: Official RoboColosseum leaderboard
| # | Entry | Score ± | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | gpt-6-astra-droid | 1,925 1852.4 to 2000.1 | rig-v1 | Official leaderboard | |
| 2 | molmoact2-droid | 1,869 1824.1 to 1918 | d8c1abd8a27d8e859455bbe514df2bcc617db0fb | Official leaderboard | |
| 3 | g05-droid | 1,744 1693.6 to 1799.3 | e312be81e90c56a55bcb26b57429bd39a335b449 | Official leaderboard | |
| 4 | claude-opus-5-5-droid | 1,636 ±92.1 | rig-v1 | Official leaderboard | |
| 5 | gr00t-n17-droid | 1,350 ±70.5 | 05e7cc97e40dbd33b0890c35cc0214fcb0547ab5 | Official leaderboard | |
| 6 | pi05-droid | 1,285 1198.6 to 1368.3 | 72824c0a93f00ce5bb8bedb7feb58953ba1da364 | Official leaderboard | |
| 7 | lingbot-vla-v2-6b-droid | 1,233 1119.6 to 1336.9 | 11c703bf6a5c1f45b3b69168482da11fdbba53d7 | Official leaderboard | |
| 8 | grok-4.7-droid | 1,160 879.8 to 1378.7 | 4.7 | Official leaderboard |
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Robotics
- ERQA In use 400 multiple-choice visual questions from robot and egocentric footage on spatial, trajectory, action and state reasoning. Accuracy.
Gemini Robotics ER 2 78.5%
- SimplerEnv In use Real-to-sim replicas of Google Robot and WidowX/Bridge setups. Policies scored by success rate on pick, move, drawer and Bridge tasks. Praxis-1 80%
- RoboLab In use Zero-shot DROID Franka policies in Isaac Lab on 120 tabletop tasks at three instruction levels, 10 episodes each. Success rate. Praxis-1 46.2%
- RoboDojo In use Generalist manipulation policies run 42 Isaac Sim tasks and 18 remote real-robot tasks. Scored by task score and success rate. GPT-6 Astra 27.6%
- Progress classification (Google DeepMind) In use Models assign each frame of a robot video to one of five task-progress bands from 0 to 100 percent. Scored by accuracy.
Gemini Robotics ER 2 57.4%
- Human proximity (Google DeepMind robotics) In use Models judge from robot camera views whether a human is within 1 meter, a spatial awareness check for safe robot behavior. Scored by accuracy.
Gemini Robotics ER 2 93%
Data audit
All 8 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access Public tasks, plus a private test set.
- Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
- Pass Independent runs Scores from the official leaderboard (8 entries).
- Pass Models scored 8 entries on its leaderboard, none a model we track.
- Note Newest score No scores yet.
- Note Frontier models No models scored yet.
- Note Headroom Scored as a rating, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on robocolosseum.ai: 8 entries at the last read, none a model we track.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.