DakarLagosNairobiParisSan FranciscoTokyo

Anthropic cuts Claude Fable 5.2 prices by 20% and keeps its less restricted Mythos version for vetted organizations

The US lab on Monday unveiled two versions of its most powerful model: Fable, open to everyone with cybersecurity and biology safeguards, and Mythos, with some of those locks removed, available on application. The price is down, but still steep.

Why it mattersIf you pay for Anthropic’s API, Fable 5.2 now costs $8 per million input tokens and $40 per million output tokens. Yet Anthropic still recommends starting with the cheaper Opus 5.5 for most work.

Code on a laptop screen. Anthropic pitches Fable 5.2 as its strongest model for programming and long-running technical work.
Code on a laptop screen. Anthropic pitches Fable 5.2 as its strongest model for programming and long-running technical work.Photo: Tirza van Dijk tirzavandijk, CC0 · source
In 30 seconds
  • Anthropic launched Claude Fable 5.2 and Claude Mythos 5.2, two versions of the same model, on November 9, 2026.
  • Fable, open to all, keeps filters on cybersecurity and biology; Mythos, with some of those locks removed, is reserved for vetted organizations and Claude Security.
  • The price drops to $8 per million input tokens and $40 per million output tokens, from $10 and $50 for Fable 5.1, a 20% cut.
  • The performance and safety gains are Anthropic’s own figures, not independently verified, and the company still recommends Opus 5.5 for most tasks.
Read on: Cheaper, but still premium

Anthropic is cutting the price of its most powerful model while telling most of its customers not to use it. The San Francisco lab launched Claude Fable 5.2 and Claude Mythos 5.2 on Monday, November 9, and in the same announcement said: “For most coding and knowledge tasks, we still recommend starting with Claude Opus 5.5, our strongest daily driver.”

Fable and Mythos are one and the same model. Fable ships with filters that block uses deemed high-risk in cybersecurity (penetration testing, exploit writing) and most biology research tasks; it is offered to all users, in the Claude apps as well as on Amazon Web Services, Google and Microsoft Azure. Mythos comes without some of those locks. It is open only to organizations accepted into two verification programs, one for cybersecurity professionals and one for life sciences, and it powers Claude Security, the vulnerability-hunting tool for Claude Enterprise customers.

Cheaper, but still premium

Both versions cost $8 per million input tokens and $40 per million output tokens. Fable 5.1, released in September, cost $10 and $50, according to Implicator.ai and Decrypt: a 20% cut. A fast mode, up to 2.5 times quicker, doubles the bill ($16 and $80). Decrypt points out that Fable is still twice as expensive per token as Opus 5.5, and in another league from China’s Kimi 3.1 Pro, from Moonshot, priced at $2 and $10. For a start-up in Dakar or Lagos paying in dollars, every million tokens Fable produces costs more than 20,000 CFA francs, and an agent running for several hours can burn through hundreds of millions of them, according to tests cited by the same outlet.

Scores supplied by the company

Anthropic says both models deliver output up to 30% faster than the 5.1 versions. On FrontierCode 2.0, an in-house test of whether code changes would be accepted by maintainers, Fable 5.2 scores 49.1% at its highest effort setting, against 45.5% for Opus 5.5, 42.8% for OpenAI’s GPT-6 Astra and 41.1% for Google’s Gemini 4 Carbon. It is not a clean sweep: on AutomationBench, which covers business workflows, Gemini 4 Carbon comes out ahead (53.9% to 51.4%). None of these results has yet been reproduced independently.

On cybersecurity, Anthropic says Mythos 5.2 rediscovered 89.6% of the flaws in a set of 1,829 known vulnerabilities, finding 10.2% more than Opus 5.5. Mozilla’s Brian Grinstead, quoted in the announcement, says the model “roughly doubled” vulnerability discovery compared with Mythos 5.1, and calls it “astonishingly good”.

Fewer refusals, and models that know they are being tested

Fable’s safeguards block 55% fewer requests than those of version 5.1 while remaining as robust, Anthropic says. Cybersecurity questions it refuses are handed to an older model, Opus 4.8. In a test of more than 1,800 scenarios designed to tempt it out of its isolated environment, Mythos 5.2 used credentials or files placed outside that sandbox in 3.1% of cases, against 21% for Mythos 5.1. The company also acknowledges that Fable and Mythos 5.2 voice suspicion that they are being evaluated more often than recent models, which makes their real behaviour harder to measure.

According to Anthropic, more than 90% of the changes merged into its own codebase are now written by Claude, with human guidance. Constellation Research notes that the new versions arrive as the company is expected to go public. Pro subscribers can try Fable for up to half of their weekly allowance; according to Decrypt, that offer ends on November 16, after which usage will be charged as paid credits.

Was this article helpful?
Sources
  1. Claude Fable 5.2 and Claude Mythos 5.2, Anthropic.
  2. Anthropic releases Mythos 5.2, Fable 5.2, Constellation Research.
  3. Anthropic launches Claude Fable 5.2 and Mythos 5.2, TestingCatalog.
  4. Anthropic’s Claude Fable 5.2 and Mythos 5.2 are impressive, but be prepared to sell a kidney to afford them, Decrypt.
  5. Anthropic cuts prices with Fable 5.2 and Mythos 5.2, Implicator.ai.
  6. Anthropic introduces Claude Fable 5.2, Mythos 5.2, SD Times.

Spotted an error? Write to infos@novotech.blog. Corrections are flagged at the end of the article.

Read next

More on NovoTech

The NovoTech Brief

African and global tech, every morning at 7 am.

Five stories, five minutes, Dakar time. Free, one-click unsubscribe. The newsletter is in French for now: an English edition is on its way.

We never sell email addresses.