Anthropic · model
Claude Mythos 5.1 benchmark scores
As of 2026-11-22, Claude Mythos 5.1 (Anthropic) has measured scores on 2 benchmarks and estimated scores on 19 more.
| Benchmark | Score | Source |
|---|---|---|
| GDPval-AA | 55.3% | estimated ± 7.6 pp, low confidence |
| Terminal-Bench 2.1 | 88.2% | estimated ± 6.5 pp, low confidence |
| BrowseComp | 88.2% | estimated ± 3.0 pp, medium confidence |
| CyberGym | 86.0% | estimated ± 10.0 pp, medium confidence |
| AA Agentic Index | 45.4% | estimated ± 8.0 pp, low confidence |
| Gert Labs | 69.7% | estimated ± 7.7 pp, medium confidence |
| AA-AnalystAgent | 44.2% | estimated ± 5.5 pp, low confidence |
| MazeBench | 5.3% | estimated ± 3.7 pp, high confidence |
| Terminal-Bench 4.1 | 60.5% | estimated ± 4.6 pp, high confidence |
| CWE-bench v1 | 57.8% | estimated ± 4.5 pp, medium confidence |
| Terminal-Bench 4.0 | 60.9% | measured |
| Terminal-Bench-Science 0.1 | 57.9% | estimated ± 14.2 pp, low confidence |
| AA AutomationBench | 68.2% | estimated ± 4.8 pp, medium confidence |
| GDP.pdf | 24.0% | estimated ± 7.4 pp, low confidence |
| AA Terminal-Bench 4.0 | 57.1% | estimated ± 6.3 pp, medium confidence |
| CWE-bench | 55.9% | estimated ± 8.2 pp, low confidence |
| MKA | 44.9% | estimated ± 3.1 pp, high confidence |
| Hermes Bench | 73.6% | estimated ± 1.5 pp, medium confidence |
| Hermes Index | 48.7% | estimated ± 5.5 pp, low confidence |
| ExploitGym | 23.4% | measured |
| Agents' Last Exam | 51.7% | estimated ± 9.6 pp, medium confidence |