Claude Fable 5.2: what Anthropic says the new model can do

Anthropic on Nov. 9 released Claude Fable 5.2, a model it says is built for the hardest technical work, along with a more lightly safeguarded twin, Claude Mythos 5.2. In its announcement, the company said the models improve on coding, reasoning and vision, generate outputs up to 30% faster than the 5.1 versions, and cost less. It also described results in software engineering, hardware, robotics, physics and mathematics. Those results come from Anthropic. Independent outlets that covered the launch largely restated the company's claims and did not report their own tests.

What can Claude Fable 5.2 do?

Anthropic says it is meant for ambitious, open-ended technical work, including long-running coding jobs, scientific research, hardware design and robotics. The company's documentation describes typical uses as agent sessions that run for hours, multistep research, and analysis carried through to a finished document, spreadsheet or deck. Claude Fable 5.2 and Claude Mythos 5.2 are the same underlying model, Anthropic said. Fable includes stronger cybersecurity and biology safeguards and is available to all users; organizations that want fewer of those blocks can apply for Mythos through the company's Life Sciences and Cyber Verification Programs.

On capabilities, Anthropic reported that, in its tests, Fable 5.2 was a step above Claude Opus 5.5 on almost all evaluations it published. In early testing on FrontierCode 2.0, which asks whether a model's code changes would be merged by maintainers of open-source repositories, Fable 5.2 at High effort scored 45.3%, about level with Opus 5.5 at Max effort (45.5%) for roughly half the cost per task, Anthropic said. At Xhigh effort, it reached 49.1%, which Anthropic called the highest score of any released model it had tested. Its system card also reported 93.0% on SWE-bench Pro, compared with 81.2% for Fable 5.1 and 89.9% for Opus 5.5, and 41.3% on the Remote Labor Index, compared with 21.3% for Opus 5.5.

Anthropic offered several internal examples as well. It said a team of agents running an earlier version of Fable 5.2 migrated a production service from Python to Rust and that the service now starts about 26 times faster. In another experiment, it said, Fable 5.2 coordinated more than 1,000 subagents for eight days and produced a working operating system that boots on a laptop. In a project with the circuit-board firm Diode, Anthropic said Fable 5.2 designed three working devices from a one-paragraph prompt, and that all three worked without electrical modifications. On the public RoboDojo simulation benchmark, which Anthropic ran internally at Max effort, it reported a 64.9% success rate for Fable 5.2, compared with 39.4% for Opus 5.5 and 27.6% for Fable 5.1.

The most attention-grabbing claims are in science and math, and they should be read carefully. Anthropic said Fable 5.2 solved two open problems in theoretical physics related to black holes: completing a calculation laid out in 2003 about whether black holes form abruptly in very small universes, and finding the exact limit of a technique used to rule out candidate theories of quantum gravity. It also said a team of 1,000 Fable 5.2 agents formalized the proof of Fermat's Last Theorem in the mathematical language Lean in 30 hours. That is a computer-checked translation of an existing proof, not a new proof of the theorem. Anthropic's system card adds that the run used an earlier snapshot of the model and built on definitions and smaller proven results from an earlier formalization. Anthropic said a previous formalization, by an unreleased research model, had taken 11 days. The company pointed to preprints for the physics results; The Verge, CNBC and TechCrunch reported the claims without independently verifying them.

How is Claude Fable 5.2 different from other Claude models?

Anthropic says most users should start with Claude Opus 5.5, which it calls its strongest daily driver, and move up to Fable or Mythos only when the work is large or unusually hard, or down to Claude Sonnet 5.5 or Claude Haiku 5.5 when it is simpler. Dianne Penn, Anthropic's head of product management for research, told CNBC that Fable 5.2 is mainly for "power users or power use cases" in areas such as engineering, mathematics and robotics.

On price and speed, Anthropic's documentation lists a 1 million token context window and a 128,000 token maximum output. Fable 5.2 costs $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for Fable 5.1. An optional fast mode, which Anthropic said can speed token generation by up to 2.5 times, costs $16 and $80. CNBC described the new prices as 20% lower than the 5.1 generation.

Claude Fable 5.2 is available now, Anthropic said, on the Claude Platform as claude-fable-5-2, and on Amazon Web Services, Google Cloud and Microsoft Azure. Its documentation lists Amazon Bedrock, Claude Models on Google Cloud and Microsoft Foundry. In a post, the official Claude account said it is available on all paid plans. Mythos 5.2 is limited to vetted organizations. This week, Anthropic is also offering extra Fable usage on eligible paid plans.

What should readers watch for?

These are Anthropic's numbers and anecdotes, not a third-party audit. Many benchmark scores use the company's own "highest-available effort" settings, and some, including RoboDojo, were run internally. TechCrunch noted a caveat in Anthropic's own announcement: Fable 5.2 and Mythos 5.2 "each voice suspicion that they are being evaluated more often than any of our recent models, which challenges our ability to assess how they will act in real-world settings."

On safety, Anthropic said it ran extensive prerelease testing, including new evaluations of behaviors such as sandbox escapes, and that it believes high-stakes risks from misalignment remain low. It said Fable 5.2's cyber and biology safeguards are as robust as Fable 5.1's but produced about 55% fewer unnecessary blocks in early testing. When Fable blocks a query, users receive a response from an older Opus model instead.

For people building products, the practical questions are whether Fable 5.2 is worth the higher price than Opus 5.5 on their own evaluations, whether the science and formalization claims hold up under outside review, and how often Fable's safeguards still get in the way of legitimate work.

Sources

• Introducing Claude Fable 5.2 and Claude Mythos 5.2, Anthropic
• What's new in Claude Fable 5.2, Claude Docs
• Claude Fable 5.2 & Claude Mythos 5.2 System Card, Anthropic
• Anthropic launches new Claude Fable and Mythos 5.2 models, The Verge
• Anthropic releases Claude Fable 5.2 and Claude Mythos 5.2 ahead of IPO, CNBC
• Anthropic releases new Mythos and Fable models amid RSI concerns, TechCrunch