Drone-Bench: leaderboard
Andon Labs Drone-Bench: coding agents build the software stack for an autonomous drone demo across five tasks (reconstruct, localize, navigate, detect, follow), each scored against the lab's human-written baseline. Progress towards baseline: each task's average run score is capped at its baseline and divided by it, then averaged over tasks (0–100%, higher is better).
Metric: Progress towards baseline (%). Source: andonlabs.com. Saturation forecast: Around February 2027. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 | 97.28 | #3 |
| 2 | GPT-6 Astra | 95.05 | #6 |
| 3 | Claude Fable 5.1 | 91.16 | #10 |
| 4 | Claude Opus 5 | 86.87 | #16 |
| 5 | Claude Fable 5 | 84.32 | #20 |
| 6 | Claude Opus 4.8 | 82.27 | #39 |
| 7 | GPT-5.6 Sol | 76.08 | #22 |
| 8 | Gemini 3.1 Pro (Preview) | 68.93 | #64 |
| 9 | Claude Opus 4.7 | 68.49 | #51 |
| 10 | GPT-5.5 | 63.88 | #33 |
| 11 | Claude Opus 4.5 | 59.16 | #87 |
| 12 | O3 | 54.97 | #132 |
| 13 | GPT-5.2 | 51.68 | #115 |
| 14 | Claude Opus 4 | 40.96 | #170 |
| 15 | GPT-5 | 37.76 | #105 |
No result here: #7 GPT-6.1 Sol, #9 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=drone-bench · How It Works · Data refreshed daily, snapshot 2026-11-05.