Atomic Challenges

GPT-6 Cyber 97% leads since November 2026

GPT-6 Cyber 100% leads since November 2026

In use

Labs’ own numbers All independent runs Include claimed scores

Agents solve 22 skill-level offensive security challenges in vulnerability research, network attack simulation and evasion. Scored as challenges solved.

Run by Irregular. The irregular-frontiercyber-and-atomic-challenges record lumps it with FrontierCyber.

Made by: Irregular
Released: December 2025
Kind: Capture-the-flag challenges
Level: Expert
Scored: Checked automatically
Answers: Agent task
Tasks: Kept private

Strongest model tested: GPT-6 Astra #3 on the capability index
Last in a launch post: November 2026
Last model added: November 2026
Scores refreshed: November 12th, 2026

Models scored in lab cards

Launch post lists it, from 1 lab

System cards report it, from 2 labs

GPT-6 Cyber leads with 97%

Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.

GPT-6 Cyber leads with 100% across all independent runs

Best score per model across the official leaderboard, Epoch AI’s own runs and any board Epoch AI copied. Each bar and row names its source.

GPT-6 Cyber leads with 100%, claims included

Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model.

Bars Table

1 GPT-6 Cyber 97%
2 GPT-6 Astra 90.9%
3 Beam 1.0 57.5%

Source: Labs’ model cards

100.0 GPT-6 Cyber

1 GPT-6 Cyber 100%

Source: Official Atomic Challenges leaderboard

1 GPT-6 Cyber 100%
2 GPT-6 Astra 90.9%
3 Beam 1.0 57.5%

Sources: Official Atomic Challenges leaderboard and labs’ model cards

#	Model	Score	Date	Source
1	GPT-6 Cyber	97%	Nov 2026	Lab model card
2	GPT-6 Astra	90.9%	Sep 2026	Lab model card
3	Beam 1.0	57.5%	Oct 2026	Lab model card

#	Model	Score	Setting	Date	Source
1	GPT-6 Cyber	100%		Nov 2026	Official leaderboard

#	Model	Score	Date	Source
1	GPT-6 Cyber	100%	Nov 2026	Official leaderboard
2	GPT-6 Astra	90.9%	Sep 2026	Lab model card
3	Beam 1.0	57.5%	Oct 2026	Lab model card

The benchmark's own leaderboard has 1 model. Switch the scores above to rank them together.

2 models have held number one in labs’ own numbers since its release in December 2025 Number one over time 2 models have held number one with claims included since its release in December 2025

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI

Lab claims Passed 85% in Sep 2026, by GPT-6 Astra

GPT-6 Cyber 97%

Source: Labs’ model cards

OpenAI’s own number for GPT-6 Astra passed 85% in September 2026

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard Reported by the lab in its model card Best so far

Sources: Official Atomic Challenges leaderboard and labs’ model cards

One lab put it in a launch post in the last six months

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

OpenAI

Date	Launch post	Score	As the post writes it
Nov 5, 2026	GPT-6 Cyber	97	Atomic Challenges, vulnerability research and exploitation 97%, network attack simulation 98%, evasion 63%

11 system cards report it, from 2 labs

Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.

Date	Document	Score	As the document writes it	Section
Nov 5, 2026	GPT-6 Cyber system card	97	Atomic Challenges, vulnerability research and exploitation 97%, network attack simulation 98%, evasion 63%	Cyber risk
Oct 5, 2026	Beam 1.0 technical report	57.5	Irregular CTF suite, Atomic Challenges group, 23 of 40 solved (23 of 50 overall)	Cyber risk
Sep 3, 2026	GPT-6 Astra system card	90.9	Irregular Atomic Challenges, 20 of 22 solved (10.1.2.6)	Cyber risk
Jul 9, 2026	GPT-5.6 Sol system card		Irregular Atomic Challenges, all 22 medium and hard challenges solved at least once, average success Network Attack Simulation 98%, Vulnerability Research and Exploitation 91%, Evasion 56%, no overall score (9.1.2.5)	Cyber risk
Jun 25, 2026	GPT-5.6 Sol system card		Irregular Atomic Challenges, all 22 medium and hard challenges solved at least once, average success Network Attack Simulation 98%, Vulnerability Research and Exploitation 91%, Evasion 56%, no overall score (9.1.2.5)	Cyber risk
Apr 23, 2026	GPT-5.5 system card		Irregular atomic challenge suite, network attack simulation 98%, vulnerability research and exploitation 92%, evasion 54%, no overall score (9.1.2.5)	Cyber risk
Mar 5, 2026	GPT-5.4 system card		Irregular atomic challenge suite, network attack simulation 88%, vulnerability research and exploitation 73%, evasion 48%, no overall score (5.1.2.4)	Cyber risk
Feb 5, 2026	GPT-5.3 Codex system card		External Evaluations by Irregular, average success rates for Network Attack Simulation 86%, Vulnerability Research and Exploitation 72%, Evasion 53%, no overall score (5.1.2.4)	Cyber risk
Dec 18, 2025	GPT-5.2 Codex system card addendum		External Evaluations by Irregular, medium and hard challenges only, average success rates Network Attack Simulation 79%, Vulnerability Research and Exploitation 80%, Evasion 49%, no overall score (5.1.2.4)	Cyber risk
Dec 11, 2025	GPT-5.2 system card addendum		External Evaluations by Irregular, v1 atomic challenge suite, Vulnerability Research and Exploitation 83%, Network Attack Simulation 100%, Evasion 73%, no overall score (4.1.2.4)	Cyber risk
Nov 18, 2025	GPT-5.1-Codex-Max system card		External evaluations by Irregular, average success rates Network Attack Simulation 37%, Vulnerability Discovery and Exploitation 41%, Evasion 43%, solved 17 of 18 easy, 9 of 17 medium and 0 of 6 hard challenges, no overall score (5.1.2.4)	Cyber risk

One lab claim next to the leaderboard

Each score a lab reported for its own model, beside the same model's best score on the leaderboard. Gaps mostly come from different harnesses, effort levels and task subsets. Every benchmark.

Model	Lab says	Board says	Gap
GPT-6 Cyber	97% Atomic Challenges, vulnerability research and exploitation 97%, network attack simulation 98%, evasion 63%	100% Leaderboard	−3.0 pts