Just in
TechAnthropic IPO Date: Nov. 19 Pricing Assumed as Nasdaq Approves ANTAIGPT-6.2 Astra and Sol Arrive as Astra API Prices Fall 20%CryptoBullish Earnings: Q3 Adjusted Revenue Up 10%, $300 Million BuybackTechNvidia Earnings: Q3 Revenue Up 103% to $115.8 BillionTechApple Outage: iCloud Sign In Issue Continues as App Store ClearsTechARTEX AI Agents Used in Key Theft and Deepfake Attacks, Amazon Says
AI

Claude Fable 5.2 and Mythos 5.2 Launch at $8 per Million Tokens

Anthropic released Claude Fable 5.2 for all users and Mythos 5.2 for vetted organizations on Nov. 9. Both cost $8 per million input tokens and $40 per million output tokens.

USA Publishing Group4 min read
Share
Abstract digital artwork showing the Claude name and geometric shapes, for the Fable 5.2 model launch
Photo: Brecht Corbeel / Unsplash

Key takeaways

  • Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on Nov. 9, priced at $8 per million input tokens and $40 per million output tokens.
  • Fable 5.2 is open to all users, while Mythos 5.2 is limited to organizations approved through Anthropic's life sciences and cyber verification programs.
  • Anthropic said 1,000 Fable 5.2 agents formally proved Fermat's Last Theorem in Lean in 30 hours, against 11 days for an earlier research model.
In this story 5 sections
  1. Who can use Fable 5.2 and who can use Mythos 5.2
  2. Faster output, and a paid fast mode
  3. Claude Fable 5.2 benchmarks against Opus 5.5
  4. A 30 hour proof of Fermat's Last Theorem
  5. Safety testing and early reactions

Anthropic released Claude Fable 5.2 on Nov. 9, along with a restricted counterpart called Claude Mythos 5.2, and priced both at $8 per million input tokens and $40 per million output tokens. The company says in its launch announcement that the two names describe one underlying model. What separates them is access: Fable carries strong cybersecurity and biology safeguards so anyone can use it, while Mythos is held back for vetted groups doing defensive security or biomedical work.

Anthropic said the new pricing is lower than what it charged before. Anthropic is a private company and does not trade on a stock exchange.

Who can use Fable 5.2 and who can use Mythos 5.2

  • Fable 5.2 is available to all users, according to the announcement.
  • Mythos 5.2 requires approval. Organizations apply through Anthropic's expanded Life Sciences Verification Program or its Cyber Verification Program.
  • Claude Security, a product for Claude Enterprise customers, now runs on Mythos 5.2.

Despite the new release, Anthropic still recommends Claude Opus 5.5 as the place to start for most coding and knowledge tasks. It lists Claude Sonnet 5.5 and Claude Haiku 5.5 for users who want to scale down.

Faster output, and a paid fast mode

Anthropic said the 5.2 models return output up to 30% faster than Fable 5.1 and Mythos 5.1. An optional fast mode speeds up token generation by as much as 2.5 times and doubles the price, to $16 per million input tokens and $80 per million output tokens.

To show what that means in dollars, we priced a sample job of 2 million input tokens and 500,000 output tokens. At standard rates it costs $36 (2 times $8, plus 0.5 times $40). In fast mode the same job costs $72.

Claude Fable 5.2 benchmarks against Opus 5.5

The scores below are Anthropic's own results and have not been independently verified:

  • FrontierSWE v2: 69.9% for Fable 5.2, compared with 62.3% for Opus 5.5 and 56.3% for Fable 5.1.
  • FrontierCode 2.0: 49.1% at Fable's highest effort setting, against 45.5% for Opus 5.5.
  • Humanity's Last Exam (with tools): 74.2%, compared with 67.7% for Opus 5.5.
  • Remote Labor Index: 41.3%, up from 17.9% for Fable 5.1. By our math that is about 2.3 times the older model's score.
  • AutomationBench: 51.4%, ahead of Opus 5.5 at 42.5% but behind Google's Gemini 4 Carbon at 53.9%.

Fable 5.2 did not win every test in the table. On the AA-Briefcase Elo rating it scored 1785, a little under Opus 5.5 at 1800. Anthropic also said that at its High effort level Fable 5.2 scores 45.3% on FrontierCode, nearly level with Opus 5.5, at roughly half the cost per task.

A 30 hour proof of Fermat's Last Theorem

The most unusual claim in the release involves math. Anthropic said a team of 1,000 Fable 5.2 agents produced a formal proof of Fermat's Last Theorem in the Lean proof language in 30 hours, with little human involvement. The agents began from short prompts and a list of intermediate theorem statements, with no mathematical references.

An earlier attempt by an unreleased research version of Claude needed 11 days, a custom harness and human guidance, the company said. The new proof is seven times shorter and used a third as many action tokens. Given the same 1,000 agent setup, Opus 5.5 finished 26% of the work after 18 hours and made little further progress by the 30 hour mark.

Safety testing and early reactions

Anthropic said it judged the risk of unsafe behavior in high stakes situations to be low. Outside testing came from the UK AI Security Institute, Frontier Design and METR, the evaluation group behind a recent report on AI research speed.

The announcement includes comments from early users. Brian Grinstead, a distinguished engineer at Mozilla, said: "Claude Mythos 5.2 is astonishingly good." Sam Zhang, a creative technologist at Unity, said Fable 5.2 "can complete tasks that previous models struggled with."

The launch lands days after OpenAI introduced its own security focused model, covered in our story on GPT-6 Cyber. Follow more model releases on our AI page.

Claude Fable 5.2 Pricing, Access and Benchmarks