CyberGym
Can the model diagnose or complete difficult cybersecurity tasks?
Cybersecurity5 ranked models5 reported values2 reportsHigher is better
Top rankings
One row per model, using its best reported score across effort settings.
| Rank | Model | Best score | Best reported setting |
|---|---|---|---|
| 1 | DeepSeek V4.1 FlashDeepSeek | 88.1% | Reported configurationDeepSeek report1 value · 1 reportSep 10, 2026 · source |
| 2 | Claude Opus 4.8Anthropic | 83.1% | Reported configurationDeepSeek report1 value · 1 reportJul 31, 2026 · source |
| 3 | DeepSeek V4 Flash 0731DeepSeek | 76.7% | max effortDeepSeek report1 value · 1 reportJul 31, 2026 · source |
| 4 | DeepSeek V4 ProDeepSeek | 52.7% | Reported configurationDeepSeek report1 value · 1 reportJul 31, 2026 · source |
| 5 | DeepSeek-V4-FlashDeepSeek | 38.7% | Reported configurationDeepSeek report1 value · 1 reportJul 31, 2026 · source |
Effort curve
Every sourced cost-linked effort value for this exact version. Lines connect complete sweeps only.
No cost-linked effort sweep for this version.
Definition and comparison boundarypublic methodology
Can the model diagnose or complete difficult cybersecurity tasks?
Security analysis or exploitation under a controlled evaluation environment. Higher is better. The value is the percentage reported in this lab's table.
Harness, tools, prompts, attempt count, and benchmark version can materially change the result. Compare within one reporting context.
DeepSeek's official V4.1-Flash release table; no pass metric, task revision, or harness is printed, so this remains a separate provider benchmark definition.
Method / source