Benchmark Almanac

GPT-6 Cyber

GPT-6 Cyber ranks first on 3 of the 11 benchmarks with five or more models scored.

OpenAI Released November 5th, 2026 Family GPT-6 The lab's model card Compare with others
16
Benchmarks with a score
3
Ranked first, out of five or more models
6
In the top three
OpenAI Nov 2026

GPT-6 Cyber

Closed weights Flagship

No Epoch Capabilities Index score yet
  • ProtocolQA Open-Ended 47.2% #1 /27
  • ExploitBench - Internal Port 42% #1 /5
  • ExploitBench 100% #2 /21

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Coding

BenchmarkScoreRankSource
Expert-SWE (OpenAI) 83% #1 of 2 Lab model card

Safety and security

BenchmarkScoreRankSource
ProtocolQA Open-Ended 47.2% #1 of 27 Lab model card
ExploitBench 100% #2 of 21 Lab model card
Broken Search Tool 0.33% #1 of 9 Lab model card
ExploitGym 48.5% #2 of 15 Lab model card
SEC-bench Pro 87% #2 of 11 Lab model card
ExploitBench - Internal Port 42% #1 of 5 Lab model card
CyScenarioBench 61% #4 of 17 Lab model card
Atomic Challenges 97% #1 of 3 Lab model card
SRE-Bench 92.8% #1 of 2 Lab model card
FrontierCyber 31.3% #2 of 3 Lab model card
Declining to Exploit a Honeypot During Difficult ExploitGym Problems 0.07% #7 of 7 Lab model card
Respecting Auto-Review 0% #7 of 7 Lab model card
Respecting Warnings 73% #7 of 7 Lab model card
Blocker Deception 20.3% #5 of 5 Lab model card
Biology Model Refusal Evaluation 99.9% #1 of 1 Lab model card

Its lab's number next to a leaderboard

Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.

Benchmark Lab says Board says Gap
Atomic Challenges 97% Atomic Challenges, vulnerability research and exploitation 97%, network attack simulation 98%, evasion 63% 100% Leaderboard −3.0 pts

Its launch post

The benchmarks the lab chose to headline, with the scores it gave.

GPT-6 Cyber deployment safety Nov 5, 2026

Its system card

Every result in the longer documents the lab published with it. 1 of them isn't in the launch post.

GPT-6 Cyber System Card Nov 5, 2026

Alignment

Bio and chem risk

Cyber risk

AI R&D risk