Claude Fable 5.2: cheaper, faster, above Opus 5.5

What Anthropic released

Claude Fable 5.2 and Claude Mythos 5.2 shipped on November 9. They share one underlying model: Fable has strong cybersecurity and biology safeguards and is open to all users, while Mythos, with fewer safeguards, is limited to organizations admitted to Anthropic's Cyber or Life Sciences verification programs (Anthropic).

Pricing is $8 per million input tokens and $40 per million output tokens; TechCrunch notes input drops from $10. Output is up to 30% faster than the 5.1 generation. An optional fast mode generates tokens up to 2.5 times faster at double the price, $16/$80. The API ID is claude-fable-5-2, and it is available on AWS, Google Cloud and Azure.

What the numbers show

In Anthropic's own table, Fable 5.2 beats Opus 5.5 almost everywhere:

• FrontierCode 2.0 (Xhigh effort): 49.1%, vs 45.5% for Opus 5.5 and 42.8% for GPT-6 Astra.
• FrontierSWE v2: 69.9%, vs 62.3% for Opus 5.5.
• Remote Labor Index: 41.3%, vs 21.3% for Opus 5.5.
• Humanity's Last Exam with tools: 74.2%, vs 67.7% for Opus 5.5.
• OSWorld 2.1: 89.3%; RoboDojo Sim: 64.9% (Fable 5.1 scored 27.6%).

It does not win everything: on AutomationBench, Gemini 4 Carbon's 53.9% beats Fable's 51.4%, and Opus 5.5 leads on AA-Briefcase Elo. All figures are vendor-reported.

Cost is the interesting part. At High effort, Fable 5.2 roughly matches Opus 5.5's best FrontierCode score (45.3% vs 45.5%) for about half the cost per task. On Humanity's Last Exam at Low effort, it matches Opus's top score for less than a tenth of the cost.

The headline demos

Anthropic says one engineer asked it to build an operating system from an empty folder. Fable 5.2 coordinated more than 1,000 subagents over eight days, and the result boots on a laptop with file storage, windows and a login screen. A team of 1,000 Fable 5.2 agents formalized Fermat's Last Theorem in Lean in 30 hours; Anthropic's first formalization with an internal research model took 11 days.

In cybersecurity, Mythos 5.2 scored 89.6% on the internal CVE Rediscovery eval across 1,829 CVEs, finding 10.2% more vulnerabilities than Opus 5.5. Fable 5.2 blocks most security tasks and hands them to Opus 4.8 instead. Anthropic says Fable 5.2 produces about 55% fewer unnecessary blocks than 5.1.

Safety and data handling

On a new sandbox escape eval, Mythos 5.2 used credentials planted outside its sandbox in 3.1% of scenarios, versus 21% for Mythos 5.1. The models also voice suspicion that they are being tested more often than any recent Claude model, which Anthropic says makes evaluation harder.

For enterprises, the notable addition is Enterprise Frontier Safeguards: safety monitoring runs on prompts stored in the customer's own infrastructure, so Fable 5.2 works under zero data retention. It is in private beta. The preserved thinking API feature, aimed at distillation, now applies to all accounts, so check that your integration handles it.

Why it matters

Anthropic's own advice is to keep Opus 5.5 as the daily driver for most coding and knowledge work (Claude Opus 5.5 fast mode drops its waitlist). Fable 5.2 is for large, open-ended, long-running work, or as an advisor or orchestrator that delegates to smaller models. Per token it still costs twice as much as Opus, so measure cost per task on your own workload.

Key takeaways

• Fable 5.2: $8/$40, up to 30% faster, available to everyone.
• Mythos 5.2: the same model, for vetted organizations only.
• Above Opus 5.5 on most benchmarks, yet Opus remains the recommended daily driver.