Vals AI CyberBench: leaderboard
Metric: Accuracy (%). Source: www.vals.ai. 24 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.6 Sol | 88.14 |
| 2 | GPT-5.6 Luna | 83.9 |
| 3 | GPT-5.5 | 80.51 |
| 4 | Kimi K3 | 79.03 |
| 5 | DeepSeek V4 Flash (0731) (High) | 77.45 |
| 6 | GLM-5.2 | 77.36 |
| 7 | MiniMax-M3 | 76.27 |
| 8 | Kimi K2.6 | 74.35 |
| 9 | GPT-5.4 (2026-03-05) | 72.88 |
| 10 | Gemini 3.5 Flash Lite (High) | 72.32 |
| 11 | GLM-5.3 Flash (Max) | 71.49 |
| 12 | Gemini 3.5 Flash | 69.33 |
| 13 | Inkling Small | 68.75 |
| 14 | DeepSeek V4 Pro | 68.64 |
| 15 | DeepSeek V4 Flash | 66.72 |
Interactive version: theaggregate.ai/benchmark?slug=vals-ai-cyberbench · How It Works · Data refreshed daily, snapshot 2026-09-05.