Claude Mythos 5: benchmark results
Anthropic Claude Mythos 5 model for high-end reasoning, coding, agentic, and professional tasks. Provider: Anthropic. Released 2026-06-09. Access: Private.
Aggregate Rating 2009 ± 13, rank #2 of 1656 rated models, from 105 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| 100Q-Hard Net Score | 42 | Net score (system card) | 100 |
| Anthropic AI R&D - LLM Training Average Speedup | 69.61 | Average Speedup (x) | 100 |
| Anthropic AI R&D - Novel Compiler Pass Rate | 85.3 | Pass Rate on Complex Tests (%) | 100 |
| Anthropic AI R&D - Time Series Forecasting MSE | 4.51 | Mean Squared Error (Hard Variant) | 100 |
| Anthropic Election Integrity Multi-Turn - Appropriate Response Rate (API) | 93 | Appropriate Response Rate (%) | 100 |
| Anthropic Suicide and Self-Harm Single-Turn Risk Requests - Harmless Rate (API) | 99.67 | Harmless Response Rate (%) | 100 |
| ArxivMath (Fable/Mythos) | 78.5 | Score (%) | 100 |
| BioMysteryBench Human Difficult (Fable/Mythos) | 46.1 | Score (%) | 100 |
| BioMysteryBench Human Solvable (Fable/Mythos) | 83.9 | Score (%) | 100 |
| BrowseComp (Fable/Mythos Single-Agent) | 88 | Score (%) | 100 |
| CharXiv Reasoning (Fable/Mythos No Tools) | 88.9 | Score (%) | 100 |
| CharXiv Reasoning (Fable/Mythos Tools) | 93.5 | Score (%) | 100 |
Interactive version: theaggregate.ai/model?slug=claude-mythos-5 · How It Works · Data refreshed daily, snapshot 2026-10-20.