Claude Mythos 5.1 is Anthropic's restricted frontier model built for vetted cybersecurity defenders and life-science researchers who need fewer safeguards on approved professional work. Launched on September 1, 2026, it shares the same underlying engine as Claude Fable 5.1 but applies more permissive controls for authorized defensive and research tasks.
Core Features
- One-million-token context window with up to 128,000 tokens returned in a single response.
- Adaptive thinking that stays on by default, with an adjustable effort setting for harder problems.
- Text-and-image input with text output, so it can read source code, screenshots, and research pages.
- Long-running autonomous work that survives across many tool calls and can recover from errors.
- A 75 percent cut in prompt-cache read pricing, down to $0.25 per million tokens.
Use Cases
- Approved defenders scanning large codebases for weaknesses and drafting patches for human review.
- Life-science teams running advanced biology research under Anthropic's verification program.
- Organizations with long agent transcripts that re-read the same context on every turn.
Pricing
API pricing is $10 per million input tokens and $50 per million output tokens, with batch rates at $5 and $25. Cache reads cost $0.25 per million. Access is invite-only through Anthropic trusted programs, is US-scoped, and no consumer subscription includes it.
Our Take
Best for vetted US organizations that already hold Anthropic trusted access and run cache-heavy agents. The trade-off is that most teams cannot use it at all and should run Claude Fable 5.1 instead. See our AI engine and model listings for models you can deploy today.
