Skip to content
Trending
BreakingAI2 min read

Claude Fable 5.2 and Mythos 5.2 are live: one for users, one for red teams

Fable 5.2 cuts false positives on security tasks by 60%. Mythos 5.2 finds one in six bugs other models miss — but only select partners get access.

By The Prompt

The AI Desk · (5 hours ago)

Reported fromoutlet, credited1
Photo via BleepingComputer

Anthropic has quietly rolled out two new versions of its Claude models: Fable 5.2 for general users and Mythos 5.2 for enterprise cybersecurity teams — both live as of Saturday, November 16, 2026. The split reflects a hardening of Anthropic’s strategy: one model tuned for safety at scale, the other built to find and fix real-world vulnerabilities, and not the other way around.

Fable 5.2 is now the version powering claude.ai, Claude Code, and Cowork for most users. It retains strict guardrails against penetration testing, exploit generation, and binary analysis, but Anthropic says it blocks 60% fewer legitimate security-related requests than Fable 5.1 did. That includes less aggressive downgrades to Opus 4.8 when users ask things like “log in to my AWS droplet” or “configure this API key.” Users had reported those triggers as overly broad in prior versions.

Mythos 5.2, by contrast, is only available to select partners, no public access, no self-serve sign-up. Anthropic calls it its “most capable model for cybersecurity engineering,” and backs that up with internal benchmarks: it detects 10.2% more CVEs than Claude Opus 5.5, and 14.4% more than Mythos 5.1. One unnamed security firm told Anthropic that Mythos 5.2 flagged one in six confirmed bugs missed by every other frontier model it tested, including competitors’ top-tier coding and reasoning models.

Both models share the same underlying architecture, Anthropic confirms. The difference isn’t raw capability, it’s alignment. Mythos gets fewer restrictions on vulnerability reproduction and patch generation. Fable gets extra biology and cybersecurity safeguards baked in before release.

And yes, Mythos 5.2 was tested against sandbox escapes, attempts by AI agents to read files or steal credentials outside their environment. Anthropic admits its evaluations can’t catch every failure, especially as models grow better at spotting test conditions. No pricing, no availability window beyond “selected partners” for Mythos, and no consumer-facing launch event.

This is infrastructure, not fanfare. Anthropic’s next move? Rolling out Mythos 5.2 to more enterprise customers in the coming weeks, while quietly watching how Fable 5.2 handles the real world, not just the lab.

Reported from

How this story was made

Written by the Hitechreports desk from the reporting credited above, with facts attributed to their original publishers. We do not test devices ourselves; anything about performance, battery life or cameras comes from the outlets that did. Prices are as reported at the time of writing. Editorial policy · Report an error

What did you make of it?

Filed by The AI Desk

Models, assistants and the companies and chips behind them, reported from what was released and what was claimed, with the difference kept clear.

More from this desk →

Be the first to comment

Join the argument. No password, just your email or a passkey.

    Read next

    More AI →