SEC-Bench ProBrowse 296

SEC-Bench Pro

Can the model diagnose or complete difficult cybersecurity tasks?

Latest stableGPT-6 System Card · result metric as reported
Cybersecurity4 ranked models4 reported values1 reportsHigher is better

Top rankings

One row per model, using its best reported score across effort settings.

4
RankModelBest scoreBest reported setting
1GPT-6 AstraOpenAI85.4%Reported configurationOpenAI report1 value · 1 reportSep 3, 2026 · source
2GPT-5.6 SolOpenAI79.1%max effortOpenAI report1 value · 1 reportSep 3, 2026 · source
3GPT-6 SolOpenAI66.3%Reported configurationOpenAI report1 value · 1 reportSep 22, 2026 · source
4GPT-6 LunaOpenAI34.2%Reported configurationOpenAI report1 value · 1 reportSep 22, 2026 · source

Effort curve

Every sourced cost-linked effort value for this exact version. Lines connect complete sweeps only.

0
No cost-linked effort sweep for this version.
Definition and comparison boundarylimited 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.

OpenAI reports GPT-6 Astra 85.4%, GPT-5.6 Sol at Max 79.1%, GPT-6 Sol 66.3%, and GPT-6 Luna 34.2%. The card does not specify the harness or effort for the GPT-6 Sol/Luna rows.

Method / source