Skip to content
What is Claude Fable 5.2? Anthropic's model for the hardest work
Back to Blog
#AI models#Claude Fable 5.2#claude-fable-5-2#Claude Fable 5.2 pricing#Fable 5.2 vs Opus 5.5#Claude Mythos 5.2

What is Claude Fable 5.2? Anthropic's model for the hardest work

Uzair bin Haroon

Claude Fable 5.2 is Anthropic's most capable generally available model, released on 9 November 2026 for long, hard technical work: large code migrations, multi-agent jobs that run for hours, and expert-level analysis. It costs $8 per million input tokens and $40 per million output tokens, twice the price of Claude Opus 5.5. Anthropic itself recommends starting with Opus 5.5 and moving up to Fable only where Opus falls short.

Use it for the jobs where a cheaper model keeps failing or needs constant correction. Do not make it the default for chat, routing or simple extraction: you pay twice the Opus rate for work Opus or Sonnet 5.5 already does.

What changed from Fable 5.1?

It is cheaper, faster and better at agent work. The price fell from $10 and $50 per million tokens to $8 and $40. Anthropic says it generates output up to 30% faster and finishes hard tasks with fewer tokens at Medium effort. Its safety classifiers also block about 55% fewer legitimate requests than Fable 5.1 did.

On Anthropic's own benchmarks, the biggest gains are in agentic coding, business workflows and long remote-work tasks:

Fable 5.2Opus 5.5Fable 5.1
Price per million tokens (in / out)$8 / $40$4 / $20$10 / $50
FrontierSWE v2 (agentic coding)69.9%62.3%56.3%
AutomationBench (business workflows)51.4%42.5%31.4%
Remote Labor Index (knowledge work)41.3%21.3%17.9%
OSWorld 2.1 (computer use)89.3%87.2%83.5%

These are Anthropic's scores, mostly at each model's highest effort, and not yet confirmed independently. Fable 5.2 does not win everything: on AutomationBench, Gemini 4 Carbon scores 53.9%. Test it on your own tasks before you switch.

Fable 5.2 or Opus 5.5?

Opus 5.5 for most work; Fable 5.2 where the task is long, ambiguous or has already beaten Opus. The token price is double, but the cost per task need not be. On FrontierCode 2.0, Anthropic reports Fable 5.2 at High effort scoring about level with Opus 5.5 at Max effort (45.3% against 45.5%) for roughly half the cost per task, because it needs fewer tokens to get there.

In practice, that gives you three ways to use it:

  • Switch up on failure. Run Opus 5.5 first and send only the jobs it fails to Fable 5.2.
  • Turn effort down. The API default is High. Anthropic shows Medium and even Low matching Opus 5.5's best scores on some tests, at a fraction of the cost.
  • Use it as advisor or orchestrator. Anthropic describes Fable 5.2 advising Haiku, Sonnet or Opus on hard steps, or handing simple subtasks down to them, so most tokens go through the cheaper models.

If Opus 5.5 is your comparison point, the Opus 5.5 pricing breakdown shows what a typical job costs there.

What will it refuse?

Most security and biology work. Fable 5.2 runs classifiers that block penetration testing, exploit generation, binary vulnerability research and most research biology. A blocked request is answered by an older model instead: Opus 4.8 for cybersecurity, Opus 5 for biology. If your product does security scanning or life-sciences research, you need Claude Mythos 5.2, the same underlying model without those safeguards, which Anthropic grants through its Cyber Verification and Life Sciences Verification Programs.

Check data handling too. Because misuse can show up only across many requests, Anthropic keeps Fable 5.2 traffic for 30 days for safety monitoring. Eligible enterprise customers can keep zero data retention, and a private beta called Enterprise Frontier Safeguards runs that monitoring inside your own infrastructure.

How do you get it?

It is live now. The API model id is claude-fable-5-2 on the Claude Platform, Microsoft Foundry and Google Cloud, and anthropic.claude-fable-5-2 on Amazon Bedrock. It reads up to 1 million tokens of context and writes up to 128,000 tokens per response. Batch processing halves the price. Fast mode, which generates up to 2.5 times faster, costs twice the standard rate: $16 and $80 per million tokens.

If you want Fable 5.2 working inside your product or workflow, with routing to cheaper models, effort settings and cost limits set up properly, see what we build or tell us what you need.

Sources