Truace
Subscribe
42/100

Emerging evidence

Cite
More
Export CSVExport JSONEmailLinkedInX

Anthropic launches Claude Fable 5.2 and Mythos 5.2, its most advanced AI models for coding, science, and cybersecurity

Evidence type
News report
Sources
1 sourceJournalism
Published
2 min read
Truvace record
TRV-2026-2466cite or check it

Anthropic has launched Claude Fable 5.2 and Claude Mythos 5.2, two new AI models that the company says can tackle some of the hardest problems in software engineering, scientific research, and cybersecurity. The announcement includes extraordinary claims: the AI solved two longstanding problems in black hole physics, helped design three working electronic devices, and coordinated

Anthropic launches Claude Fable 5.2 and Mythos 5.2, its most advanced AI models for coding, science, and cybersecurity

The gain

Anthropic says Mythos 5.2 identified 89.6% of tested vulnerabilities on its internal CVE Rediscovery evaluation versus 81.3% for Claude Opus 5.5, and Fable 5.2 scored 49.1% on FrontierCode 2.0 versus 42.8% for GPT-6 Astra.

In brief

Anthropic launched Claude Fable 5.2 and Claude Mythos 5.2 on November 9, 2026. The company says Mythos 5.2 identified 89.6% of tested vulnerabilities on its internal CVE Rediscovery evaluation, against 81.3% for Claude Opus 5.5 and 78.3% for Mythos 5.1.

Anthropic also reports that Fable 5.2 scored 49.1% on FrontierCode 2.0, ahead of 42.8% for OpenAI's GPT-6 Astra and 41.1% for Google's Gemini 4 Carbon. Both models generate output up to 30% faster than the 5.1 versions and cost $8 per million input tokens and $40 per million output tokens.

All figures come from Anthropic's own testing, and the article notes they do not prove equal performance in real-world security settings. Gemini 4 Carbon led the reported AutomationBench business workflow test, and Anthropic acknowledges testing cannot predict every failure after deployment.

Main points

  1. Fable 5.2 at High effort matched Claude Opus 5.5 at Max effort on FrontierCode 2.0 at roughly half the cost per task, according to Anthropic's early testing.
  2. Anthropic says Claude now writes more than 90% of the changes merged into its core codebase, with human engineers providing guidance and oversight.
  3. An earlier version of Fable 5.2 migrated a production service from Python to Rust, with startup about 26 times faster, Anthropic says.
  4. Fable 5.2 coordinated more than 1,000 subagents over eight days to build a prototype operating system that boots on a laptop.
  5. Fable 5.2 designed three hardware devices built by manufacturing partner Diode; all three worked without electrical modifications.
  6. 1,000 Fable agents formalized the existing proof of Fermat's Last Theorem in 30 hours, which is verification of a known proof rather than a new discovery.
  7. Mythos 5.2 is restricted to vetted organizations for cybersecurity and sensitive biological research; Fable 5.2 is broadly available with restrictions.

The full story

On coding, Anthropic says Fable 5.2 at Xhigh effort achieved the highest score among released models it evaluated on FrontierCode 2.0. Medium effort stays the default in Claude apps and Claude Code. In the operating-system experiment, the model started from an empty folder and handled coding, testing and code reviews through its subagents, producing a kernel, disk and network drivers, a graphical interface, a login screen and an installer.

In science, Anthropic says Fable 5.2 completed calculations on two open problems involving black holes and quantum gravity and linked the results to research preprints so researchers can examine them. The Fermat's Last Theorem formalization translated an existing proof into a computer-checkable format, which the article says could cut the time mathematicians spend checking difficult arguments.

Fable 5.2 is available through Anthropic's platforms and AWS, Google Cloud and Microsoft Azure. An optional fast mode runs up to 2.5 times faster at twice the token prices. Anthropic says external organizations including the UK AI Security Institute and METR evaluated the models, and its internal findings suggest reduced rates of several unwanted behaviors.

Talk it overFour questions for a class, a team or the dinner table

  1. Who gets this benefit, and who was left out of the study or story?

    The gain Anthropic says Mythos 5.2 identified 89.6% of tested vulnerabilities on its internal CVE Rediscovery evaluation versus 81.3% for Claude Opus 5.5, and Fable 5.2 scored 49.1% on FrontierCode 2.0 versus 42.8% for GPT-6 Astra.

  2. How do we know? What can this kind of evidence show, and what can’t it?

    The evidence This rests on a news report, from TechStartups.com.

  3. What does it leave open? What else would you want to know before acting on it?

    The record’s own caveat The benchmark results are Anthropic-reported, not proof that the model will perform equally well across real-world security environments, and Anthropic acknowledges that testing cannot predict every failure after deployment.

  4. Have you seen AI used this way where you study, work or live? Did it go the same way?

The questions are the same on every record. The details under them are taken from this record as written.

Sources

1 on file
  1. TechStartups.comtechstartups.com
    Journalism2026-11-10

The debate

More from Science