Benchmarks / Coding

Drone-Bench

Models write drone-surveillance code across five tasks (reconstruction, localization, navigation, detection, following); progress toward the human-authored baseline.

11models

–Open Index weight

–last updated

Andon Labspublisher

#	Model	Effort	Score	Cost / task	Harness
1	Opus 5.5	max	97.3%	–	–
2	GPT-6 Astra	max	95.1%	–	–
3	Fable 5.1	max	91.2%	–	–
4	Opus 5	max	86.9%	–	–
5	Fable 5	max	84.3%	–	–
6	Opus 4.8	max	82.3%	–	–
7	Gemini 3.1	–	68.9%	–	–
8	Opus 4.5	high	59.2%	–	–
9	o3	high	55.0%	–	–
10	o1	high	31.9%	–	–
11	GPT-4o	high	14.0%	–	–