AI Prompt
BlogAbout
← Back

Claude Fable 5.2, Mythos 5.2, and More

5 min read

Anthropic released two new models today: Claude Fable 5.2 and Claude Mythos 5.2. They are the same underlying model. Fable ships with strong cybersecurity and biology safeguards, which is why it's available to everyone. Mythos drops some of those safeguards and is limited to organizations approved through Anthropic's Cyber and Life Sciences Verification Programs.

The pricing is the first thing I noticed. Both cost $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for the 5.1 models. Anthropic says they generate output up to 30% faster than Fable 5.1 and Mythos 5.1. There's also an optional fast mode that speeds up generation by up to 2.5x for $16 input and $80 output per million tokens. Developers can use claude-fable-5-2 on the Claude Platform today, and Fable 5.2 is also on AWS, Google Cloud, and Microsoft Azure. (Anthropic)

The benchmark table puts Fable 5.2 ahead of Opus 5.5 on almost everything Anthropic tested. On FrontierCode 2.0, which checks whether a model's code changes would be merged by open-source maintainers, Fable 5.2 scores 49.1% at Xhigh effort, against 45.5% for Opus 5.5 and 42.8% for GPT-6 Astra. FrontierSWE v2 comes in at 69.9%, Humanity's Last Exam with tools at 74.2%, and OSWorld 2.1 at 89.3%. The biggest jump is robotics: 64.9% on RoboDojo Sim, against 27.6% for Fable 5.1. The win isn't universal. Gemini 4 Carbon edges it on AutomationBench, 53.9% to 51.4%.

The demos are ambitious. Anthropic says a team of about 1,000 Fable 5.2 agents formalized the proof of Fermat's Last Theorem in Lean in 30 hours. In the same setup, Opus 5.5 finished 26% of the formalization in 18 hours. Fable 5.2 also designed three working electronic devices from a one-paragraph prompt, solved two open problems in black hole physics, and, in one experiment, coordinated more than 1,000 subagents over eight days to build a prototype operating system that boots on a laptop.

Anthropic is clear that this isn't a daily driver. Its advice is to start with Opus 5.5 for most coding and knowledge work, then move up to Fable or Mythos for the hardest, most open-ended problems. Fable 5.2 can also act as an Advisor that smaller models consult, or as an Orchestrator that hands simpler tasks to them. Medium effort is the default in the Claude apps and Claude Code, and at High effort Anthropic says Fable 5.2 roughly matches Opus 5.5's best FrontierCode score for about half the cost per task.

What Else Happened

Mythos 5.2 is Anthropic's strongest cyber model. On Anthropic's internal CVE Rediscovery eval, run against 1,829 CVEs, Mythos 5.2 found 10.2% more vulnerabilities than Opus 5.5 and 14.4% more than Mythos 5.1. Claude Enterprise customers can get it through Claude Security without joining the verification program, and Claude Security now watches repos as code changes and connects through an API and MCP. (Anthropic)

Fable's safety filters block less often, and blocked requests fall back to older models. In early testing, Fable 5.2 triggered about 55% fewer blocks than Fable 5.1. When a request is blocked, cyber requests get an answer from Opus 4.8 and biology requests from Opus 5. That matters if you build on Fable and need to know which model actually answered. (Anthropic)

Anthropic deliberately held back AI R&D capabilities. Anthropic says it limited training gains in areas such as accelerator kernel optimization and LLM architecture work, so Fable 5.2 performs about at Opus 5.5's level there. That is an unusual thing for a lab to say in a launch post. (Anthropic)

Preserved thinking now applies to every API account. As part of its anti-distillation work, Anthropic is rolling preserved thinking out to all accounts, which stops callers from editing Claude's earlier context to pull out its reasoning. If your integration rewrites earlier turns, check the docs. (Anthropic)

Mythos 5.2 specs are on the docs page. It has a 1M-token context window, 128K max output, and adaptive thinking that's always on with high as the default effort. It's on the Claude API, Amazon Bedrock, Google Cloud, and Microsoft Foundry, with retirement no sooner than November 9, 2027. (Claude docs)

Paid Claude plans get extra Fable usage this week. Subscribers on eligible plans get one to three Fable resets based on how long they've subscribed, and Pro plans and standard Team seats can use Fable for up to 50% of weekly usage limits. (Anthropic)

What I'm Reading

That's it for today. 👋