FLY-EVAL++ - Single-Step Prediction - Safety Constraints: leaderboard
FLY-EVAL++ (arXiv 2609.04021v1, Table 1) single-step flight-state prediction (S1), D4 safety score: FAA-grounded safety constraint library with penalties. Deterministic evidence-based verification of 45-field JSON predictions; percent, higher is better. Only the 21 of 66 screened models above 90% protocol compliance are listed.
Metric: S1 D4 safety score (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Grok 4 | 92.8 | #189 |
| 2 | Gemini 3 Pro | 92.7 | #82 |
| 3 | GPT-5 | 92.1 | #105 |
| 4 | DeepSeek R1 0528 | 91.2 | #232 |
| 5 | O4 Mini | 90.4 | #185 |
| 6 | Kimi K2 (Thinking) | 88 | #252 (Kimi K2) |
| 7 | Gemini 2.5 Pro | 86.2 | #159 |
| 8 | GLM-4.6 | 78.8 | #265 |
| 9 | Claude Sonnet 4.5 | 70.6 | #148 |
| 10 | Llama 3.1 405B | 70.2 | #476 |
| 11 | Claude 3.7 Sonnet (20250219) | 69.9 | #209 |
| 12 | Kimi K2 0905 | 68.5 | #292 |
| 13 | DeepSeek V3.1 | 67.5 | #275 |
| 14 | DeepSeek V3.2 Exp | 66.4 | #241 |
| 15 | DeepSeek V3 | 64.5 | #329 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=fly-eval-plus-plus-single-step-prediction-safety-constraints · How It Works · Data refreshed daily, snapshot 2026-11-05.