Benchmark Almanac

RoboColiseum

HexaShell 98.5% tops its leaderboard Claude Opus 5.5 76.3% leads the models we track
Not in model cards

AgiBot Genie robot in Genie Sim 3.0 (Isaac Sim): instruction, robustness, manipulation and spatial boards after post-training on provided sim data.

Instruction board crowded (many pi0.5 fine-tunes 0.72-0.84). Manipulation and spatial not saturated. Many noisy third-party rows. Organizer baselines mixed with submissions.

Made by
AgiBot
Released
January 2026
Kind
Simulated robots
Level
Everyday
Scored
Checked automatically
Answers
Agent task
Strongest model tested
GPT-6 Astra #3 on the capability index
Last model added
October 2026
Scores refreshed
November 12th, 2026
31
Entries on its own leaderboard, 9 models we track

HexaShell tops the leaderboard with 98.5%

Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. 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. The best model we track is Claude Opus 5.5, at 76.3%.

31 entries. Scroll sideways for the rest.

  1. 98.5 HexaShell
  2. 96.6 MXI SimWorld v0 Preview
  3. 90.2 FF Omnimind
  4. 82.8 PyRUA-Lean
  5. 78.8 GO-3 Nano Preview
  6. 77.5 Prime-0
  7. 76.9 HiDream-O1-Embodied
  8. 76.3 Claude Opus 5.5
  9. 76.0 GPT-6 Astra
  10. 70.9 GPT-6.1 Sol
  11. 70.8 DM0.5
  12. 66.9 Gemini 4 Argon
  13. 65.0 GigaBrain-0.7
  14. 60.2 KaneRSI
  15. 59.8 WNDFC
  16. 59.4 Xiaomi-Robotics-1
  17. 58.4 OpenWAM-Alpha
  18. 57.9 pi0.5_refine
  19. 57.4 π0.5
  20. 56.7 kp-robot
  21. 56.3 ACoT-VLA
  22. 55.8 lingbot-vla-v2
  23. 55.1 DeepSeek V4.1 Flash
  24. 53.6 RoboMan-Pi05
  25. 47.2 Dream-1 SFT 500 steps
  26. 46.8 GR00T N1.7
  27. 46.7 InternVLA-A1.5
  28. 43.0 Xiaomi-Robotics-0
  29. 42.3 Muse Wonder
  30. 33.5 Dream-1 zero-shot
  31. 28.9 π0
  1. 1 HexaShell 98.5%
  2. 2 MXI SimWorld v0 Preview 96.6%
  3. 3 FF Omnimind 90.2%
  4. 4 PyRUA-Lean 82.8%
  5. 5 GO-3 Nano Preview 78.8%
  6. 6 Prime-0 77.5%
  7. 7 HiDream-O1-Embodied 76.9%
  8. 8 Claude Opus 5.5 76.3%
  9. 9 GPT-6 Astra 76%
  10. 10 GPT-6.1 Sol 70.9%
  11. 11 DM0.5 70.8%
  12. 12 Gemini 4 Argon 66.9%
  13. 13 GigaBrain-0.7 65%
  14. 14 KaneRSI 60.2%
  15. 15 WNDFC 59.8%
  16. 16 Xiaomi-Robotics-1 59.4%
  17. 17 OpenWAM-Alpha 58.4%
  18. 18 pi0.5_refine 57.9%
  19. 19 π0.5 57.4%
  20. 20 kp-robot 56.7%
  21. 21 ACoT-VLA 56.3%
  22. 22 lingbot-vla-v2 55.8%
  23. 23 DeepSeek V4.1 Flash 55.1%
  24. 24 RoboMan-Pi05 53.6%
  25. 25 Dream-1 SFT 500 steps 47.2%
  26. 26 GR00T N1.7 46.8%
  27. 27 InternVLA-A1.5 46.7%
  28. 28 Xiaomi-Robotics-0 43%
  29. 29 Muse Wonder 42.3%
  30. 30 Dream-1 zero-shot 33.5%
  31. 31 π0 28.9%

Source: Official RoboColiseum leaderboard

#EntryScoreSettingDateSource
1 HexaShell 98.5% Nov 2026 Official leaderboard
2 MXI SimWorld v0 Preview 96.6% Nov 2026 Official leaderboard
3 FF Omnimind 90.2% Oct 2026 Official leaderboard
4 PyRUA-Lean 82.8% Nov 2026 Official leaderboard
5 GO-3 Nano Preview 78.8% Oct 2026 Official leaderboard
6 Prime-0 77.5% Oct 2026 Official leaderboard
7 HiDream-O1-Embodied 76.9% Sep 2026 Official leaderboard
8 Claude Opus 5.5 76.3% Sep 2026 Official leaderboard
9 GPT-6 Astra 76% Sep 2026 Official leaderboard
10 GPT-6.1 Sol 70.9% Sep 2026 Official leaderboard
11 DM0.5 70.8% Sep 2026 Official leaderboard
12 Gemini 4 Argon 66.9% Sep 2026 Official leaderboard
13 GigaBrain-0.7 65% Aug 2026 Official leaderboard
14 KaneRSI 60.2% Nov 2026 Official leaderboard
15 WNDFC 59.8% Aug 2026 Official leaderboard
16 Xiaomi-Robotics-1 59.4% Oct 2026 Official leaderboard
17 OpenWAM-Alpha 58.4% Oct 2026 Official leaderboard
18 pi0.5_refine 57.9% Jul 2026 Official leaderboard
19 π0.5 57.4% Apr 2025 Official leaderboard
20 kp-robot 56.7% Jul 2026 Official leaderboard
21 ACoT-VLA 56.3% Jun 2026 Official leaderboard
22 lingbot-vla-v2 55.8% Oct 2026 Official leaderboard
23 DeepSeek V4.1 Flash 55.1% Sep 2026 Official leaderboard
24 RoboMan-Pi05 53.6% Nov 2026 Official leaderboard
25 Dream-1 SFT 500 steps 47.2% Oct 2026 Official leaderboard
26 GR00T N1.7 46.8% Apr 2026 Official leaderboard
27 InternVLA-A1.5 46.7% Sep 2026 Official leaderboard
28 Xiaomi-Robotics-0 43% Aug 2026 Official leaderboard
29 Muse Wonder 42.3% Oct 2026 Official leaderboard
30 Dream-1 zero-shot 33.5% Oct 2026 Official leaderboard
31 π0 28.9% Oct 2024 Official leaderboard

3 models have held number one on the official leaderboard since its release in January 2026

Lab claims count from the day the card came out, independent runs from the model’s release.

Physical Intelligence OpenAI Anthropic
Own leaderboard 3 leaders
Claude Opus 5.5 76.3%

Source: Official RoboColiseum leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Claude Opus 5.5 π0.5 π0 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Claude Opus 5.5 π0.5

Source: Official RoboColiseum leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Robotics

All 103 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%
  • 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 29%
  • RoboLab In use Zero-shot DROID Franka policies in Isaac Lab on 120 tabletop tasks at three instruction levels, 10 episodes each. Success rate. FLUX 3 Action 42.9%
  • 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%
  • Moment finding (Google DeepMind) In use Models find the moment in a robot video when an event happens. Scored by accuracy and mean absolute time distance in seconds. Gemini Robotics ER 2 91.3%