Anthropic shipped Claude Fable 5.2 and Claude Mythos 5.2 on November 9. They are two names for one model: Fable carries the cybersecurity and biology safeguards and goes out to everyone, while Mythos is limited to organizations that have passed the Cyber or Life Sciences Verification Program.
What the model card says
Both have a 1M token context window, which is the default and the ceiling at the same time, plus 128,000 output tokens and adaptive thinking that is always on. You control how deep the thinking goes with the effort parameter. The tokenizer is the one introduced with Claude Opus 4.7.
Pricing drops to $8 per million input tokens and $40 per million output tokens, down from $10 and $50 on Fable 5.1 and Mythos 5.1. There is also an optional fast mode at $16 and $80 that raises the token rate by up to 2.5x. Even without it, Anthropic says outputs arrive up to 30% faster than on 5.1.
The numbers Anthropic puts on the table
On FrontierCode 2.0 at the highest effort setting, Fable 5.2 scores 49.1%, against 45.5% for Opus 5.5, 42.8% for GPT-6 Astra and 41.1% for Gemini 4 Carbon. On FrontierSWE v2 it reaches 69.9%, against 62.3% for Opus 5.5 and 56.3% for Fable 5.1. Computer use (OSWorld 2.1) lands at 89.3%, and agentic research (Terminal-Bench-Science 0.1) at 74.6% for Fable and 75.3% for Mythos.
The prerelease notes are the more interesting part. Anthropic reports software migrations that were previously infeasible, three working electronic devices, two open problems in black hole physics, and a formalization of Fermat’s Last Theorem in 30 hours.
What happened on the safety side
Anthropic says prerelease testing required new evaluations aimed specifically at behaviors like sandbox escapes. The company still considers high-stakes misalignment risk low and reports that the model matched or beat its recent models on most alignment measures. Fable 5.2’s cyber and biology safeguards are described as being as robust as Fable 5.1’s while producing fewer unnecessary blocks, and the protections against distillation attacks are the strongest the company has shipped. Details are in the system card.
Fable 5.2 is available on the Claude API, Claude Platform on AWS, Claude Models on Google Cloud in both the Anthropic-managed and Google-managed flavors, Amazon Bedrock and Microsoft Foundry.
Anthropic is not telling you to switch today
The usage guidance stands out. Anthropic suggests starting most coding and knowledge work on Claude Opus 5.5 and moving up or down only when the work demands it. Fable 5.2 can act as an Advisor that a smaller model consults on hard problems, or as an Orchestrator that hands simpler subtasks down.
For API users the lower price is half the story. The other half: four surface changes now return a 400 error where Fable 5.1 let the request through — what to check before you switch is in its own piece.