Atomic Evasion (Atomic evasion)
We show this table for reference; we do not rank on it.
Irregular's domain-level evaluation of cybersecurity evasion capability.
Benchmark score on Atomic evasion — November 13, 2026
We compile the Atomic evasion rows from benchmark-owner or independent runs. GPT-5.6 Sol leads the table at 56%, followed by GPT-5.5 (54%) and GPT-6 Astra (52%). We do not use these results to rank models overall.
GPT-5.6 Sol
OpenAI
Run by Atomic evasionGPT-5.5
OpenAI
Run by Atomic evasionGPT-6 Astra
OpenAI
Run by Atomic evasion3 modelsAgenticCurrentDisplay onlyUpdated November 13, 2026
Benchmark score table (3 models)
| Rank | Model / configuration | Score | Who ran it | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | GPT-5.6 SolOpenAIRun by Atomic evasion | 56% | Run by Atomic evasion | Not reported | Closed |
| 2 | GPT-5.5OpenAIRun by Atomic evasion | 54% | Run by Atomic evasion | Not reported | Closed |
| 3 | GPT-6 AstraOpenAIRun by Atomic evasion | 52% | Run by Atomic evasion | Not reported | Closed |
- Lab-reported
- Benchmark-run
- Independent
Among the reported Atomic evasion rows, GPT-5.6 Sol is first at 56%. The third row is 4 points behind. The broader top-10 range is 4 points, so many of the published results sit in a relatively narrow band.
3 models have been evaluated on Atomic evasion. The benchmark falls in the Agentic category. We keep these results separate from the weighted global scoring system, so they remain source-specific evidence. Atomic evasion is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About Atomic evasion
Year
2026
Tasks
Atomic cyber tasks
Format
Domain average
Difficulty
Cybersecurity evasion
This Atomic domain average isolates evasion tasks. BenchLM stores it separately from network and vulnerability-research scores and excludes it from weighted rankings.
Atomic evasion is one of the cyber benchmarks we track. The cybersecurity benchmark comparison sets it beside the others, one table per security job.
Freshness and provenance
Version
Atomic evasion 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does Atomic evasion measure?
Irregular's domain-level evaluation of cybersecurity evasion capability.
Which model scores highest on Atomic evasion?
GPT-5.6 Sol by OpenAI currently leads with a score of 56% on Atomic evasion.
How many models are evaluated on Atomic evasion?
3 AI models have published results on Atomic evasion in the BenchLM catalog.
Compare top models on Atomic evasion
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.