Claude Fable 5.2 pricing: API costs and subscription plans

👁 114 reads

On the evening of November 9, I opened two Anthropic price lists side by side: the one for its API and the one for subscription plans. Claude Fable 5.2 costs $8 per million input tokens and $40 per million output tokens, 20% less than Fable 5.1. The headline rates are simple, but the cost of a specific task also depends on prompt caching, Batch API, fast mode and whether you pay for API usage or a subscription.

Here’s the breakdown. Anthropic lists its prices in US dollars. Whether VAT applies and which currency appears on your invoice depends on your account and company details.

Claude Fable 5.2 API pricing per million tokens

Claude Mythos 5.2 has exactly the same listed rates as Fable 5.2, but is available only to organizations verified by Anthropic. A few details in the documentation can affect your bill more than the headline rates:

• The same price applies to the entire context. The context window is one million tokens, with no surcharge for long prompts. A 900,000-token prompt costs the same per token as a 9,000-token prompt.
• Cache reads are inexpensive. For Fable and Mythos, a cache read costs 2.5% of the input price. For most Anthropic models, it is 10%; for Opus 5.5 and Sonnet 5.5, it is 5%. If you repeatedly send the same long document or code, caching can make a significant difference.
• US processing costs more. The inference_geo: "us" parameter, which keeps processing in the United States, multiplies all rates by 1.1.
• Fast mode cannot be combined with Batch API. Cache discounts and the multiplier for processing in the US do apply to fast-mode prices.
• Thinking is always on. Fable 5.2 uses adaptive thinking, which cannot be disabled. Reasoning tokens are billed as output tokens, and their number can be adjusted with the effort parameter. The default API level is high.

What one task costs in six scenarios

To put the rates in context, consider a typical programming task: you send 50,000 tokens of code and instructions and receive a 5,000-token response, including reasoning. At Fable 5.2’s standard API rates, the input costs 50,000 × $8 / 1,000,000 = $0.40, and the output costs 5,000 × $40 / 1,000,000 = $0.20. That’s $0.60 in total.

Prompt caching makes for the most interesting comparison. Suppose 45,000 of the 50,000 input tokens are the same code you send again within five minutes. The first request pays $0.45 to write those tokens to the cache (45,000 × $10 / 1,000,000), plus $0.04 for 5,000 new input tokens and $0.20 for the response: $0.69 altogether. Each subsequent request pays just $0.009 to read the cached tokens (45,000 × $0.20 / 1,000,000), plus the same $0.04 for new input and $0.20 for the response. That comes to $0.249 instead of $0.60. The savings grow with each request in a longer session working on the same repository.

At 1,000 such tasks a month, the total without caching is $600. The same number of tasks on Opus 5.5 would cost $300, while Fable 5.1 would cost $750. Fable 5.2 is therefore cheaper than its predecessor, but not cheap: it still costs twice as much as Opus 5.5. Anthropic says that in some tests, Fable at a lower effort level matches Opus’s best result at a fraction of the per-task cost - for example, on FrontierCode 2.0 at high effort, at roughly half the cost. That is the company’s claim, which I’ll check in my own test.

My test results

I checked the per-task cost in my own test at medium effort. The test used Polish prompts and included code, puzzles, images and text. Across 27 tasks, Fable 5.2 cost $0.75 and Opus 5.5 cost $0.52. Fable used fewer output tokens (16,434 versus 23,839), so it cost only 1.44 times as much per task, despite its higher price per token. GPT-6.1 Sol completed the same set for $0.11 and Gemini 4 Argon for $0.14. All models scored 100% on code, puzzles and questions about images.

Which Claude subscription plans include Fable?

There’s a launch promotion, but it comes with conditions. Pro and standard Team seats get one week of Fable access included, from November 9 through November 16 at 11:59 p.m. Pacific Time. During that period, Fable can use no more than half the weekly limit. These plans also get one free limit reset, available until the same deadline. Max, Ultra and premium seats get one to three resets depending on subscription history: one for less than six months of paid subscription, two for six months to a year, and three for more than a year. These resets expire on March 8, 2027, at 11:59 p.m. Pacific Time.

There’s one eligibility requirement: the paid plan must have been active before November 9 at 10:00 a.m. Pacific Time. The Free plan, self-serve Enterprise and usage-billed Enterprise are excluded from the promotion. A reset doesn’t increase the limit; it restores usage already spent. Anthropic says Fable consumes limits faster than other models. After the promotional week, Pro users pay for Fable using usage credits if they enable them. Claude Code fast mode on every plan uses those credits only and does not count against subscription limits.

When does Claude Fable 5.2 make sense?

The numbers point to a straightforward split. If you use Claude in the app and want Fable only for harder tasks, Pro’s promotional week should be enough to see whether the difference is worth paying for. If you work in Claude Code on large projects every day, the Max plan at $100 a month, with half the limit available for Fable, could cost less than paying for every API token. Anthropic does not say how many tokens fit into that limit, though, so I can’t calculate a direct comparison. For API use, Fable makes sense where Opus 5.5 falls short, ideally with prompt caching or Batch API. More launch details are available in this Claude Fable 5.2 guide.

Frequently asked questions

How much does Claude Fable 5.2 cost?

API pricing is $8 per million input tokens and $40 per million output tokens. Fable is included in the Max plan at $100 a month and the Ultra plan at $200 a month, up to half the weekly limit.

Is Fable 5.2 cheaper than Fable 5.1?

Yes, by 20%. Fable 5.1 cost $10 per million input tokens and $50 per million output tokens; Fable 5.2 costs $8 and $40. Cache writes and reads are cheaper too.

Is Claude Fable 5.2 free with the Pro plan?

Only during the first week, through November 16, 2026, at 11:59 p.m. Pacific Time, and only up to half the weekly limit. After that, Pro users pay for Fable using usage credits.

How much does Claude Mythos 5.2 cost?

It costs the same as Fable 5.2: $8 per million input tokens and $40 per million output tokens. The difference is access: Mythos is available only to organizations verified by Anthropic.

Sources

• Claude documentation: Pricing (checked November 9, 2026)
• Claude documentation: Claude Fable 5.2 (checked November 9, 2026)
• Claude subscription pricing (checked November 9, 2026)
• Claude Help Center: Fable 5.2 launch promotion
• Claude Code documentation: Fast mode
• Anthropic: Introducing Claude Fable 5.2 and Claude Mythos 5.2 (November 9, 2026)

// read moreClaude Fable 5.2: How to Use It in Claude Code and the APIGuidesClaude Fable 5.2 Pricing & API: Costs, Cache, Batch, Model IDGuidesClaude Fable 5.2 vs. Mythos 5.2: What’s the Difference?AI Tools

Kinetyra