Benchmark Almanac

Robo-bench

GPT-6.1 Sol 28.3% leads since October 2026
Not in model cards

General agents control a simulated 7-DoF Franka arm and a 29-DoF humanoid through motor commands, sensors and camera images only, across 24 tasks.

The agent gets the robot datasheet and kinematic model but never the MuJoCo simulator itself. Most tasks are held out. The board had seven models on October 13th, 2026.

Made by
Vals AI
Released
October 2026
Kind
Simulated robots
Level
Expert
Scored
Checked automatically
Answers
Agent task
Tasks
Kept private
Strongest model tested
GPT-6 Astra #3 on the capability index
Last model added
October 2026
Scores refreshed
November 12th, 2026
7
Models scored on its own leaderboard

GPT-6.1 Sol leads with 28.3%

Best score per model from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best.

  1. 28.3 GPT-6.1 Sol
  2. 23.3 GPT-6 Astra
  3. 14.6 Claude Opus 5.5
  4. 10.4 Muse Wonder Full
  5. 5.83 Claude Fable 5.1
  6. 0.83 Gemini 4 Argon
  7. 0.21 Grok 4.7
  1. 1 GPT-6.1 Sol 28.3%
  2. 2 GPT-6 Astra 23.3%
  3. 3 Claude Opus 5.5 14.6%
  4. 4 Muse Wonder Full 10.4%
  5. 5 Claude Fable 5.1 5.83%
  6. 6 Gemini 4 Argon 0.83%
  7. 7 Grok 4.7 0.21%

Source: Official Robo-bench leaderboard

#ModelScoreSettingDateSource
1 GPT-6.1 Sol 28.3% Sep 2026 Official leaderboard
2 GPT-6 Astra 23.3% Sep 2026 Official leaderboard
3 Claude Opus 5.5 14.6% Sep 2026 Official leaderboard
4 Muse Wonder Full 10.4% Oct 2026 Official leaderboard
5 Claude Fable 5.1 5.83% Sep 2026 Official leaderboard
6 Gemini 4 Argon 0.83% Sep 2026 Official leaderboard
7 Grok 4.7 0.21% Sep 2026 Official leaderboard

Accuracy vs. whole-run API cost

Estimated API cost covers all 24 tasks, not one task. The line joins the runs that no other run beats on both measures. We redrew this chart from the numbers on the official leaderboard.

0% 10% 20% 30% $0$10$20$30 API cost for 24-task run ($), lower is better Accuracy Claude Fable 5.1: Accuracy 5.83%, api cost for 24-task run $20.4 GPT-6 Astra: Accuracy 23.3%, api cost for 24-task run $28.5 Grok 4.7: Accuracy 0.21%, api cost for 24-task run $11.3 Claude Opus 5.5: Accuracy 14.6%, api cost for 24-task run $11.4 Gemini 4 Argon: Accuracy 0.83%, api cost for 24-task run $17.3 Muse Wonder Full: Accuracy 10.4%, api cost for 24-task run $19.7 GPT-6.1 Sol: Accuracy 28.3%, api cost for 24-task run $4.97 GPT-6.1 Sol 0% 10% 20% 30% $0$10$20$30 API cost for 24-task run ($), lower is better Accuracy Claude Fable 5.1: Accuracy 5.83%, api cost for 24-task run $20.4 GPT-6 Astra: Accuracy 23.3%, api cost for 24-task run $28.5 Grok 4.7: Accuracy 0.21%, api cost for 24-task run $11.3 Claude Opus 5.5: Accuracy 14.6%, api cost for 24-task run $11.4 Gemini 4 Argon: Accuracy 0.83%, api cost for 24-task run $17.3 Muse Wonder Full: Accuracy 10.4%, api cost for 24-task run $19.7 GPT-6.1 Sol: Accuracy 28.3%, api cost for 24-task run $4.97 GPT-6.1 Sol
ModelSettingAccuracyAPI cost for 24-task run
GPT-6.1 Sol 28.3% $4.97
GPT-6 Astra 23.3% $28.48
Claude Opus 5.5 14.6% $11.38
Muse Wonder Full 10.4% $19.68
Claude Fable 5.1 5.8% $20.4
Gemini 4 Argon 0.8% $17.33
Grok 4.7 0.2% $11.25

Number one over time

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

OpenAI
Own leaderboard 1 leader
GPT-6.1 Sol 28.3%

Source: Official Robo-bench 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% Near the ceiling (85%) GPT-6.1 Sol GPT-6 Astra Claude Fable 5.1 0% 20% 40% 60% 80% 100% Near the ceiling (85%) GPT-6.1 Sol Claude Fable 5.1

Source: Official Robo-bench 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 104 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%