Claude Fable 5.2 pricing, what changed, and whether a small business should pay for it

The short answer: Claude Fable 5.2 is Anthropic's most capable model on general sale, released on 9 November 2026. It costs $8 per 1M input tokens and $40 per 1M output tokens, down from $10 / $50 for Fable 5.1 (Anthropic pricing). It ranks first on the independent Artificial Analysis index. But it still costs twice as much as Claude Opus 5.5 ($4 / $20), and Anthropic itself tells developers to start with Opus 5.5 for most work. For most small-business jobs, Opus 5.5 or Sonnet 5.5 is the better buy. Fable 5.2 is worth paying for on long, multi-step agent work where a cheaper model keeps falling short.

Every number below comes from Anthropic's own pages or from Artificial Analysis, an independent benchmark site, and is labelled with who measured it. Checked on 10 November 2026. Prices are list prices per 1 million tokens on Anthropic's API.

Claude Fable 5.2 pricing

Sources: Anthropic pricing and models overview. The API model id is claude-fable-5-2. Batch processing is half price on every model. Fast mode, which Anthropic says is up to 2.5x faster on Fable 5.2, doubles the rate.

In the Claude apps, the Claude pricing page says Fable is included on Max with up to half your weekly limit, and on Pro it is paid with usage credits.

What changed from Fable 5.1

It got cheaper. Every rate fell by 20%: input from $10 to $8, output from $50 to $40, and cache reads from $0.25 to $0.20. Fable 5.1 is not retired. Anthropic lists it as a legacy model that is still available, so nothing breaks if you leave it running, but there is no good reason to stay on it.

It got stronger. On Artificial Analysis's Intelligence Index, Fable 5.2 scores 65 at max effort, ranked first of 74 models listed on 10 November 2026. On the same day, Claude Opus 5.5 scores 59 and Fable 5.1 scores 49 (Fable 5.2 on Artificial Analysis). Anthropic reports 69.9% on the FrontierSWE v2 coding benchmark, against 62.3% for Opus 5.5 and 56.3% for Fable 5.1 (Anthropic-reported, from Anthropic's Fable page).

It thinks more by default. Fable 5.2's default effort setting is high, against medium for Opus 5.5, per the models overview. Thinking is billed as output tokens, so on the same task Fable 5.2 may write more output than Opus 5.5 as well as charge more per token. That makes the gap per finished job larger than the rate card alone.

Claude Mythos 5.2 arrived with it. Anthropic says Fable 5.2 and Mythos 5.2 are the same underlying model. Mythos 5.2 does not have the safeguards that Fable carries in risky areas, and it is only available to vetted cyberdefenders and life scientists through Anthropic's access programs (Mythos page). It costs the same. A typical business uses Fable 5.2 and cannot buy Mythos.

Fable 5.2 vs Opus 5.5: what a job costs

The table prices three example jobs at list price. The token counts are our own illustrations of typical small-business work, not measurements, and they leave out caching and thinking tokens. Use them to compare the models with each other, not to predict your exact bill.

How we worked it out: for the email job, 2 million input tokens at $8 is $16, plus 0.5 million output tokens at $40 is $20, so $36 on Fable 5.2. The other cells use the same sum with each model's rates.

The pattern holds at any volume. Fable 5.2 costs twice Opus 5.5 and four times Sonnet 5.5 for the same tokens. If your job repeats the same long instructions or documents, cache reads cut the input side sharply on every model, and Fable 5.2's cache rate is the same $0.20 per 1M as Opus 5.5's.

Is Claude Fable 5.2 worth it for a small business?

Pay for Fable 5.2 if:

• You run long agent tasks, such as a multi-hour coding job or a research task with many steps, and Opus 5.5 at higher effort still fails your own test. That is the case Anthropic recommends Fable 5.2 for.
• A wrong answer is expensive and the token bill is small next to it. A one-off contract review costs about a dollar more on Fable 5.2 than on Opus 5.5.
• You are on Fable 5.1 today. Switching to claude-fable-5-2 cuts the bill by 20% and gets you a stronger model.
• You already pay for Claude Max, where Fable is included up to half your weekly limit, so there is no extra cost to try it.

Do not pay for Fable 5.2 if:

• Your work is drafting, customer email, social posts, summaries or everyday coding. Sonnet 5.5 at $2 / $10 or Opus 5.5 at $4 / $20 does these well, and the saving adds up quickly at volume.
• You run high-volume sorting or routing. Claude Haiku 5.5 or another low-cost model is the right tool.
• You have not tested a cheaper model on your real task. A higher score on an index does not mean a better result on your job.

How to try it

1. Pick one real task you already run, with an input you can reuse.
2. Run it on Opus 5.5, then on Fable 5.2, by changing only the model id. Keep the effort setting the same.
3. Compare the result quality and the actual bill in the Claude Console, not the rate card.
4. Move only the tasks where Fable 5.2 clearly wins. Leave everything else on the cheaper model.

For every current model, with prices and sources, see our AI model tracker. For a side-by-side page, see Claude Fable 5.2 vs Claude Opus 5.5. If you want a plan that picks the right model for each job in your business, that is what a Dapols plan does.