Pre-Flight: leaderboard
Zero-shot Inspect multiple-choice accuracy on 300 aviation operational knowledge items, dataset revision 439d2d1. Only models rerun on this revision are comparable; the paper also lists older unrepeated scores separately. Provider defaults and local quantization labels are retained.
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around January 2027. 31 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 82.7 | #31 |
| 2 | GPT-5 | 80.3 | #98 |
| 3 | Gemini 2.5 Pro | 79 | #152 |
| 4 | Claude Opus 4.8 | 79 | #37 |
| 5 | GPT-5.1 | 77 | #135 |
| 6 | GPT-5.4 | 75.7 | #79 |
| 7 | Gemini 3.5 Flash | 75.7 | #60 |
| 8 | Gemini 2.5 Flash | 75 | #245 |
| 9 | Claude Sonnet 4.5 | 74.7 | #142 |
| 10 | Llama 3.3 70B | 74.7 | #579 |
| 11 | Llama 4 Scout | 74.3 | #656 |
| 12 | Gemma 4 26B A4B | 73.3 | #358 |
| 13 | Claude Sonnet 4.6 | 72.7 | #90 |
| 14 | Claude Haiku 4.5 | 72.3 | #280 |
| 15 | GPT-OSS-120B | 70.7 | #341 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=pre-flight · How It Works · Data refreshed daily, snapshot 2026-10-20.