Claude Fable 5.2 Is Out: $8 per Million Tokens, Cheaper Modes Do the Job
In short
- Anthropic released Claude Fable 5.2 on November 9. It is available to everyone, on every major cloud.
- Price: $8 per million input tokens, $40 per million output tokens. Fast mode costs double and runs up to 2.5x faster.
- The money angle: at medium effort it matches Opus 5.5 on some tasks for about 40% of the cost, says Anthropic.
Anthropic just shipped Claude Fable 5.2. The model is stronger than Fable 5.1 on almost every benchmark Anthropic lists. The interesting part for you is not the score. It is how little you can pay for the same result.
What happened
Fable 5.2 and Mythos 5.2 are the same underlying model. Fable has safeguards for cybersecurity and biology, so anyone can use it. Mythos has no such limits and goes only to vetted organizations through Anthropic’s verification programs.
The API price is $8 per million input tokens and $40 per million output tokens. An optional fast mode costs $16 and $80 and generates tokens up to 2.5x faster. The model ID is claude-fable-5-2. It runs on AWS, Google Cloud and Azure.
On Anthropic’s own tests, agentic coding (FrontierCode 2.0) jumped from 32.4% to 49.1%. Business workflows (AutomationBench) went from 31.4% to 51.4%. Gemini 4 Carbon still scores higher there, at 53.9%. These are Anthropic’s numbers. Outside reviewers did not check the post.
How to use it
Do not run it at maximum effort by default. Anthropic’s own data says lower effort already gets you far. On coding at High effort Fable 5.2 scores 45.3%, level with Opus 5.5 at 45.5%, for about half the cost per task. On Humanity’s Last Exam at Low effort it matches Opus 5.5’s best score for less than a tenth of the cost.
Anthropic also suggests using Fable as an advisor or orchestrator. It plans, and Haiku or Sonnet do the simple steps. That is the pattern I would test first.
Do the math: one agent run with 1M input tokens and 200K output tokens costs $8 + $8 = $16 in standard mode, about €14 at roughly 0.87 € per dollar. Fast mode doubles it to about €28. Say a task costs you €10 on a bigger setting. If medium effort does the job at 40% of that, you pay €4. Run 100 tasks a month and you save €600. That is a hypothetical, built on Anthropic’s ratio. Test it on your own tasks.
What it means
Anthropic itself still recommends Opus 5.5 as the daily driver. Fable is for the hardest work. So this is not a model you switch everything to. It is a premium tool you call when a task is worth it.
One more detail: Anthropic admits Fable 5.2 and Mythos 5.2 voice suspicion of being tested more often than earlier models. That makes safety testing harder. Read the claims with that in mind.
Ivan’s Take
Worth testing, not worth switching to blindly. At $40 per million output tokens it is not cheap, and the savings only show up if you pick the effort level well. I would run it as the planner and let a small model do the boring work.
Source: Anthropic
If it doesn’t save you time or money, you won’t see it here
- Facts, not hypeEvery price and number is checked against the official source
- Always the real priceYou see what it costs before you click
- I say when it’s weakEven when it’s OpenAI or Google
