Claude Fable 5 and Claude Mythos 5 are high-performance large language models built for complex reasoning and long-horizon tasks. Developers and enterprises use these models to process massive datasets that require the 1-million-token context window. While Fable 5 is the widely available version with built-in safety classifiers, Mythos 5 offers the same raw capabilities without those filters for approved users.
These models introduce a new approach to model reasoning called adaptive thinking. Instead of returning a raw chain of thought, the models provide summarized reasoning blocks or omit them entirely based on your configuration. This makes them particularly useful for building agents that need to work through multi-step problems before delivering a final result.
Key features
- 1-million-token context window for analyzing entire codebases or libraries
- 128k output token limit per request to support long-form content generation
- Adaptive thinking mode that adjusts reasoning depth automatically
- Refusal stop reasons in Fable 5 that identify which safety classifier triggered
- Server-side fallback parameters to automatically retry requests on other models
- Support for vision, code execution, and programmatic tool calling
- 30-day data retention policy for enterprise security compliance
How to use
- Obtain an API key from the Anthropic Console or a supported cloud provider like AWS Bedrock.
- Set your model ID to claude-fable-5 or claude-mythos-5 in your request headers.
- Configure the effort parameter to determine how much reasoning the model should perform.
- Set the thinking.display parameter to summarized if you want to see the reasoning steps.
- Implement a fallback strategy to handle cases where Fable 5 refuses a specific prompt.
Use cases
- Analyzing a 1,000-page technical manual to troubleshoot a specific hardware failure
- Building a coding agent that can refactor an entire repository while maintaining context
- Summarizing weeks of meeting transcripts to identify project blockers and action items
- Processing complex financial reports to find discrepancies across multiple fiscal years
Pricing
Both models cost $10 per million input tokens and $50 per million output tokens. You are not billed for requests that Claude Fable 5 refuses before it generates any output. Check the official website for current pricing and credit options.
FAQ
What is Claude Fable 5?
It is Anthropic's most capable model designed for demanding reasoning and long-horizon agentic work.
Is Claude Fable 5 free?
No, it is a commercial model accessed via the API with a pay-as-you-go pricing structure.
What is the difference between Fable 5 and Mythos 5?
They have the same capabilities, but Mythos 5 lacks safety classifiers and is restricted to Project Glasswing members.
Does it support images?
Yes, both models include vision support for analyzing and describing visual data.
Can I see the raw thinking process?
No, these models only return summarized reasoning or omit the thinking blocks entirely.
Where can I access these models?
Fable 5 is on the Claude API, AWS Bedrock, and Google Cloud, while Mythos 5 requires special approval.
