Drone-Bench - Reconstruct: leaderboard
Drone-Bench Reconstruct: build a 3D model of an office from videos, recover each frame's pose, and supply a slicing function that yields a 2D obstacle map. Score is the average over runs of each run's best of up to ten submissions, as a percentage (higher is better); the human baseline scores 82.2 and no model has passed it.
Metric: Reconstruct score (%). Source: andonlabs.com. Saturation forecast: Around February 2027. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 | 71.03 | #3 |
| 2 | GPT-6 Astra | 67.53 | #6 |
| 3 | Claude Fable 5.1 | 51.95 | #10 |
| 4 | Claude Opus 5 | 43.19 | #16 |
| 5 | Claude Fable 5 | 38.72 | #20 |
| 6 | Claude Opus 4.8 | 33.96 | #39 |
| 7 | GPT-5.6 Sol | 33.74 | #22 |
| 8 | GPT-5.5 | 27.18 | #33 |
| 9 | O3 | 24.87 | #132 |
| 10 | Claude Opus 4 | 24.82 | #170 |
| 11 | GPT-5.2 | 23.19 | #115 |
| 12 | Claude Opus 4.5 | 18.21 | #87 |
| 13 | GPT-5 | 17.7 | #105 |
| 14 | Claude Opus 4.7 | 16.64 | #51 |
| 15 | Gemini 3.1 Pro (Preview) | 10.68 | #64 |
No result here: #7 GPT-6.1 Sol, #9 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=drone-bench-reconstruct · How It Works · Data refreshed daily, snapshot 2026-11-05.