Anthropic has introduced Claude Fable 5.1 and Claude Mythos 5.1, two models designed for coding, research, and other long-running knowledge-work tasks.
The company says both models share the same underlying capabilities, but differ in safeguards and availability. Claude Fable 5.1 is generally available to customers, while Claude Mythos 5.1 is restricted to vetted organizations participating in Anthropic’s trusted-access programs.
Anthropic positions Fable 5.1 for workloads that require extended reasoning and independent task execution. These include agentic software development, multi-step research, document, spreadsheet, and presentation work, as well as computer-use workflows.
The model supports a 1-million-token context window and up to 128,000 output tokens. A context window is the amount of text, instructions, files, and tool history a model can consider in one conversation. Anthropic also says adaptive thinking is always enabled, with developers able to adjust the model’s effort level depending on the complexity of a task.
Fable 5.1 uses the API identifier claude-fable-5-1, while Mythos 5.1 uses claude-mythos-5-1.
Anthropic says Fable 5.1 retains the same base pricing as Claude Fable 5:
- Input: USD 10 per million tokens
- Output: USD 50 per million tokens
- Cache reads: USD 0.25 per million tokens
The cache-read price is one-quarter of the previous rate. This could reduce costs for long agent sessions that repeatedly reference the same instructions or project context. Anthropic estimates that some typical or highly agentic workloads could become 25% to 45% cheaper, although actual savings will depend on context reuse, session length, and how many retries or tool calls a task requires.
The reduced cache-read price applies to both Fable 5.1 and Mythos 5.1. It does not mean Claude subscriptions have become 25% to 45% cheaper.
Mythos 5.1 remains restricted
Mythos 5.1 is positioned for advanced cybersecurity and biology research. Anthropic says access is currently limited to a small group of vetted organizations through programs under Project Glasswing.
The company says Mythos 5.1 is presently available to selected US organizations, while broader access is still being developed. Claude Security also now runs on Mythos 5.1, according to Anthropic.
Using Mythos 5.1 requires accepting a default 30-day data-retention policy for safety monitoring.
Fable 5.1 adds safeguards for wider access
Anthropic says Fable 5.1 uses the same underlying model as Mythos 5.1 but includes additional safeguards intended to make advanced capabilities suitable for general availability.
The company says Fable 5.1 can identify software vulnerabilities in source code, but its safeguards still prevent or limit activities such as penetration testing, exploit generation, and binary-based vulnerability scanning. Anthropic also says its updated biology safeguards intervene on benign requests 85% less often than those introduced with the original Fable 5. These are the company’s internal results and have not been independently validated in the announcement.
Developers should review compatibility changes
Developers moving from Claude Fable 5 should review several API changes. Fable 5.1 does not support forced tool use through tool_choice values of any or a specific tool. Anthropic recommends automatic tool selection with strict tool-use settings or structured outputs instead.
Anthropic also warns that thinking blocks are tied to the model that produced them. Applications that manually construct conversation histories should preserve those blocks correctly and avoid editing earlier turns in ways that invalidate them.
The company recommends rerunning application evaluations after migrating, particularly for agent loops, tool use, fallback behavior, token consumption, and long-running workflows.
Claude Fable 5.1 is available through Anthropic’s API and partner platforms. Mythos 5.1 remains limited to Anthropic’s trusted-access programs.
Community
0 reader comments
No comments yet. Join the conversation with a useful, respectful response.