AnthropicClaude
Claude Mythos Preview
Anthropic · Claude. Released Apr 7, 2026.
Claude Mythos Preview is a model from Anthropic in the Claude family, released Apr 7, 2026. evals.report tracks 10 reported Claude Mythos Preview benchmark scores across SWE-bench Verified, SWE-bench Pro, GPQA Diamond, Humanity's Last Exam, OSWorld, GAIA: A Benchmark for General AI Assistants, METR Task-Completion Time Horizons, CharXiv, and 2 more — each shown with its benchmark, metric, source status, and date, and never combined into a single ranking.
10 results
Benchmark results 10
Compare this model| Benchmark | Category | Score | Metric | Status | Date | |
|---|---|---|---|---|---|---|
| SWE-bench Verified | Coding | 93.9% | % resolved | Unverified | Apr 7, 2026 | Details |
| SWE-bench Pro | Coding | 77.8% | % resolved | Unverified | Apr 7, 2026 | Details |
| GPQA Diamond | Reasoning | 94.6% | accuracy | Unverified | Apr 7, 2026 | Details |
| Humanity's Last Exam | Reasoning | 56.8% | accuracy | Unverified | Apr 7, 2026 | Details |
| OSWorld | Agents | 79.6% | task success rate | Unverified | Apr 7, 2026 | Details |
| GAIA: A Benchmark for General AI Assistants | Agents | 52.3% | accuracy | Unverified | Apr 7, 2026 | Details |
| METR Task-Completion Time Horizons | Agents | 1044.8 min | 50% time horizon | Official | Apr 7, 2026 | Details |
| CharXiv | Multimodal | 93.2% | accuracy | Unverified | Apr 7, 2026 | Details |
| SWE-bench Multimodal | Coding | 59.0% | % resolved | Verified | Apr 7, 2026 | Details |
| Terminal-Bench 2.0 | Agents | 82.0% | task success | Unverified | Apr 7, 2026 | Details |