FLY-EVAL++ - Single-Step Prediction - Safety Constraints: leaderboard
Metric: S1 D4 safety score (%). Source: arxiv.org. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Grok 4 | 92.8 |
| 2 | Gemini 3 Pro | 92.7 |
| 3 | GPT-5 | 92.1 |
| 4 | DeepSeek R1 0528 | 91.2 |
| 5 | O4 Mini | 90.4 |
| 6 | Kimi K2 (Thinking) | 88 |
| 7 | Gemini 2.5 Pro | 86.2 |
| 8 | GLM-4.6 | 78.8 |
| 9 | Claude Sonnet 4.5 | 70.6 |
| 10 | Llama 3.1 405B | 70.2 |
| 11 | Claude 3.7 Sonnet (20250219) | 69.9 |
| 12 | Kimi K2 0905 | 68.5 |
| 13 | DeepSeek V3.1 | 67.5 |
| 14 | DeepSeek V3.2 Exp | 66.4 |
| 15 | DeepSeek V3 | 64.5 |
Interactive version: theaggregate.ai/benchmark?slug=fly-eval-plus-plus-single-step-prediction-safety-constraints · How It Works · Data refreshed daily, snapshot 2026-09-19.