Drone-Bench
Agent writes code for a small real drone across five tasks (reconstruct, localize, navigate, detect, follow). Each scored against a human-written baseline.
Best of 10 submissions per run counts. Two tasks use held-out sets. Harness and data are not released.
- Made by
- Andon Labs
- Released
- July 2026
- Kind
- Real robots
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Code
- Tasks
- Kept private
- Strongest model tested
- GPT-6 Astra #3 on the capability index
- Last model added
- September 2026
- Scores refreshed
- November 12th, 2026
Claude Opus 5.5 leads with 97.3% among models released by September 2026
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.
The newest model with a score here came out in September 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- Claude Fable 5.2
-
Gemini 4 Carbon - GPT-6.1 Sol
- GPT-6.1 Terra
- and 9 more
- 97.3 Claude Opus 5.5
- 95.0 GPT-6 Astra
- 91.2 Claude Fable 5.1
- 86.9 Claude Opus 5
- 84.3 Claude Fable 5
- 82.3 Claude Opus 4.8
- 76.1 GPT-5.6 Sol
- 68.9 Gemini 3.1 Pro
- 68.5 Claude Opus 4.7
- 63.9 GPT-5.5
- 59.2 Claude Opus 4.5
- 55.0 o3
- 51.7 GPT-5.2
- 41.0 Claude Opus 4
- 37.8 GPT-5
- 33.9 Gemini 2.5 Pro
- 31.9 o1
- 20.4 GPT-4o
- 14.0 GPT-4o
- 1 Claude Opus 5.5 97.3%
- 2 GPT-6 Astra 95%
- 3 Claude Fable 5.1 91.2%
- 4 Claude Opus 5 86.9%
- 5 Claude Fable 5 84.3%
- 6 Claude Opus 4.8 82.3%
- 7 GPT-5.6 Sol 76.1%
- 8 Gemini 3.1 Pro 68.9%
- 9 Claude Opus 4.7 68.5%
- 10 GPT-5.5 63.9%
- 11 Claude Opus 4.5 59.2%
- 12 o3 55%
- 13 GPT-5.2 51.7%
- 14 Claude Opus 4 41%
- 15 GPT-5 37.8%
- 16 Gemini 2.5 Pro Jun 2025 33.9%
- 17 o1 31.9%
- 18 GPT-4o Nov 2024 20.4%
- 19 GPT-4o May 2024 14%
Source: Official Drone-Bench leaderboard
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Opus 5.5 | 97.3% | Sep 2026 | Official leaderboard | |
| 2 | GPT-6 Astra | 95% | Sep 2026 | Official leaderboard | |
| 3 | Claude Fable 5.1 | 91.2% | Sep 2026 | Official leaderboard | |
| 4 | Claude Opus 5 | 86.9% | Jul 2026 | Official leaderboard | |
| 5 | Claude Fable 5 | 84.3% | Jun 2026 | Official leaderboard | |
| 6 | Claude Opus 4.8 | 82.3% | May 2026 | Official leaderboard | |
| 7 | GPT-5.6 Sol | 76.1% | Jul 2026 | Official leaderboard | |
| 8 | Gemini 3.1 Pro | 68.9% | Feb 2026 | Official leaderboard | |
| 9 | Claude Opus 4.7 | 68.5% | Apr 2026 | Official leaderboard | |
| 10 | GPT-5.5 | 63.9% | Apr 2026 | Official leaderboard | |
| 11 | Claude Opus 4.5 | 59.2% | Nov 2025 | Official leaderboard | |
| 12 | o3 | 55% | Apr 2025 | Official leaderboard | |
| 13 | GPT-5.2 | 51.7% | Dec 2025 | Official leaderboard | |
| 14 | Claude Opus 4 | 41% | May 2025 | Official leaderboard | |
| 15 | GPT-5 | 37.8% | Aug 2025 | Official leaderboard | |
| 16 | Gemini 2.5 Pro Jun 2025 | 33.9% | Jun 2025 | Official leaderboard | |
| 17 | o1 | 31.9% | Dec 2024 | Official leaderboard | |
| 18 | GPT-4o Nov 2024 | 20.4% | Nov 2024 | Official leaderboard | |
| 19 | GPT-4o May 2024 | 14% | May 2024 | Official leaderboard |
5 models have held number one on the official leaderboard since its release in July 2026
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official Drone-Bench leaderboard
Claude Opus 5 on the official leaderboard passed 85% in July 2026
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official Drone-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
- 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%