Claude Mythos Preview
Anthropic·
LLMsflagshipcloud
Overview
Anthropic's 'Capybara tier' flagship, withheld from public release due to cybersecurity risk. Available only to 12 Project Glasswing enterprise partners (incl. Amazon, Apple, Google, Microsoft, NVIDIA) via a $100M credit pool. No public API. $25/M input, $125/M output tokens.
Capabilities and innovations
Frontier reasoningAdvanced codingAgentic workflowsMathematical reasoningCapybara TierProject Glasswing Restricted AccessWithheld for Cybersecurity Risk
Benchmarks
| Benchmark | Score | Source |
|---|---|---|
| GPQA Diamond | 94% | unsourced |
| Humanity’s Last Exam (no tools) | 56.8% | unsourced |
| SWE-bench Verified | 93.9% | unsourced |
| SWE-bench Pro | 77.8% | unsourced |
| Terminal-Bench 2.x | 82% | unsourced |
- MMLU: not reported by the vendor
- MMLU-Pro: not reported by the vendor
- MMMU-Pro: not reported by the vendor
Links
More from Anthropic
Data curated by AI Model Timeline. See the methodology for admission criteria, benchmark eras and source priority.