# Atomic evasion leaderboard

external benchmark

> Every figure below is reproduced as its upstream source published it: nothing is modelled, estimated, interpolated or converted. `Unavailable` means no source published the value — it is never a zero. Each value carries its evidence state and the date it was observed.

Page: https://modelchoice.dev/leaderboards/benchmarks/external-atomic-evasion

Atomic Evasion. Every model the catalog carries a published Atomic evasion value for, ranked by that value.

## Benchmark record

| Field | Value |
| --- | --- |
| Category | external |
| Measure | Domain average |
| Tasks | Atomic cyber tasks |
| Difficulty | Cybersecurity evasion |
| Published by | [Assessing GPT-5.6 Sol](https://www.irregular.com/research/assessing-gpt-5.6-sol) |

Irregular's domain-level evaluation of cybersecurity evasion capability.

## Atomic evasion ranking

3 models with a published Atomic evasion value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.

| Rank | Model | Provider | Domain average | Evidence |
| --- | --- | --- | --- | --- |
| 1 | [GPT-5.6 Sol](https://modelchoice.dev/models/gpt-5-6-sol) | OpenAI | 56 | Observed 2026-11-21 · source benchlm:models |
| 2 | [GPT-5.5](https://modelchoice.dev/models/gpt-5-5) | OpenAI | 54 | Observed 2026-11-21 · source benchlm:models |
| 3 | [GPT-6 Astra](https://modelchoice.dev/models/gpt-6-astra) | OpenAI | 52 | Observed 2026-11-21 · source benchlm:models |

Rows are ordered by the value the source published, highest first. The source does not state whether a higher value is the better result, so this document does not either: for a benchmark that measures a rate of failure, read the table from the bottom.
