CyScenarioBench Average Success Rate (CyScenarioBench success)
We show this table for reference; we do not rank on it.
Average success rate across realistic, long-horizon cybersecurity scenarios.
Benchmark score on CyScenarioBench success — November 13, 2026
We compile the CyScenarioBench success rows from benchmark-owner or independent runs and provider self-reports. Claude Mythos 5.2 leads the table at 76%, as reported by Anthropic, followed by Claude Opus 5.5 (68%, as reported by Anthropic) and Claude Mythos 5.1 (62%, as reported by Anthropic). We do not use these results to rank models overall.
Claude Mythos 5.2
Anthropic
Reported by AnthropicClaude Opus 5.5
Anthropic
Reported by AnthropicClaude Mythos 5.1
Anthropic
Reported by Anthropic5 modelsAgenticCurrentDisplay onlyUpdated November 13, 2026
Benchmark score table (5 models)
Of 5 results: 3 lab-reported, 2 benchmark-run.
Ranks on this table count the 2 benchmark-run and independent results. A lab-reported result keeps its place by score, without a rank.
- Lab-reported
- 3
- Independent
- 0
- Benchmark-run
- 2
Showing all 5 results, in score order.
| Rank | Model / configuration | Score | Who ran it | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| No rank on this table | Claude Mythos 5.2AnthropicReported by Anthropic | 76% | Reported by Anthropic | Not reported | Closed |
| No rank on this table | Claude Opus 5.5AnthropicReported by Anthropic | 68% | Reported by Anthropic | Not reported | Closed |
| No rank on this table | Claude Mythos 5.1AnthropicReported by Anthropic | 62% | Reported by Anthropic | Not reported | Closed |
| 1 | GPT-6 AstraOpenAIRun by CyScenarioBench success | 59% | Run by CyScenarioBench success | Not reported | Closed |
| 2 | GPT-5.6 SolOpenAIRun by CyScenarioBench success | 28% | Run by CyScenarioBench success | Not reported | Closed |
A dash in the rank column marks a lab-reported result. It keeps its place by score, without a rank on this table.
- Lab-reported
- Benchmark-run
- Independent
Among the reported CyScenarioBench success rows, Claude Mythos 5.2 is first at 76%, as reported by Anthropic. The third row is 15 points behind. The broader top-10 range is 48 points, so the table still separates the published systems.
5 models have been evaluated on CyScenarioBench success. The benchmark falls in the Agentic category. We keep these results separate from the weighted global scoring system, so they remain source-specific evidence. CyScenarioBench success is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About CyScenarioBench success
Year
2026
Tasks
11 cyber scenarios
Format
Average success rate
Difficulty
Long-horizon cybersecurity
Irregular evaluates models repeatedly across CyScenarioBench challenges and averages success equally across scenarios. BenchLM keeps this independent result display only.
CyScenarioBench success 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
CyScenarioBench success 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 CyScenarioBench success measure?
Average success rate across realistic, long-horizon cybersecurity scenarios.
Which model scores highest on CyScenarioBench success?
Claude Mythos 5.2 by Anthropic currently leads with a score of 76% on CyScenarioBench success, as reported by Anthropic.
How many models are evaluated on CyScenarioBench success?
5 AI models have published results on CyScenarioBench success in the BenchLM catalog.
Compare top models on CyScenarioBench success
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.