Claude Fable 5.2 pricing calculator
What one request and one month cost, with caching, fast mode, Batch API and US data residency
Anthropic prices Claude Fable 5.2 at 8 dollars per million input tokens and 40 dollars per million output tokens. Fast mode costs 16 and 80 dollars per million, 2 times those prices, for up to 2.5 times the output speed. The Batch API costs 4 and 20 dollars per million, half the standard prices. Cache reads cost 0.20 dollars per million tokens, and US-only inference multiplies every token price by 1.1. Enter one request's token counts to see what the request and the month cost.
Updated November 17, 2026
Anthropic prices Claude Fable 5.2 (model ID claude-fable-5-2) by token type. Its pricing page lists base input at 8 dollars per million tokens (MTok), output at 40 dollars per MTok, 5 minute cache writes at 10 dollars per MTok, 1 hour cache writes at 16 dollars per MTok, and cache hits and refreshes at 0.20 dollars per MTok. The model uses adaptive thinking that is always on, so thinking tokens are billed as output tokens at 40 dollars per MTok. Enter the token counts of one request. The calculator prices the request and the month, and compares the same request on Fable 5.1 and Opus 5.5.
Formula and assumptions
- The cache write rate is the 5 minute rate or the 1 hour rate that you select.
- Fast mode and the Batch API are mutually exclusive. Anthropic's fast mode documentation says fast mode is not available with the Batch API, and the calculator shows no result when both options are on.
- Thinking tokens are billed as output tokens. Adaptive thinking cannot be disabled, so the output count you enter should include them.
- Each request in the month has the same token counts. If your requests differ, run the calculator once for each group.
- The cache write tokens that you enter are billed at the write rate on each request. If only the first request writes the cache, enter 0 here and price that first request alone.
- The Batch API option halves all four line items. Anthropic's batch documentation says all usage is charged at 50 percent of the standard API prices.
- US-only inference multiplies the final result by 1.1, matching Anthropic's statement that the multiplier applies to all token categories on Claude 4.6 and later models.
- The rates are the Claude API rates with global routing. Partner platforms are not included.
- Amazon Bedrock and Claude Models on Google Cloud (Google managed) set their own prices. On Claude Platform on AWS and Claude in Microsoft Foundry, the charges convert to Claude Consumption Units at 0.01 USD per CCU.
- The calculator checks input plus output against the 1M token context window that Anthropic lists, because all input and output tokens count toward it. It also checks the output against the 128k max output tokens.
- All sums use whole-number arithmetic on the per-token rates. Amounts below 1 USD are shown exactly. Amounts of 1 USD or more are rounded to the cent.
- Token counts are estimates. Actual usage depends on encoding, model version, and prompt structure.
This tool does not execute LLM calls. All computation happens locally in your browser.
Reference rates
Reference rates, verified 2026-11-17 against the Anthropic pricing page (https://platform.claude.com/docs/en/about-claude/pricing). Units are USD per token.
| Token type | Fable 5.2 | Fable 5.1 | Opus 5.5 |
|---|---|---|---|
| Input | 0.000008 | 0.00001 | 0.000004 |
| Output | 0.00004 | 0.00005 | 0.00002 |
| Cache read | 0.0000002 | 0.00000025 | 0.0000002 |
| Cache write, 5 minutes | 0.00001 | 0.0000125 | 0.000005 |
| Cache write, 1 hour | 0.000016 | 0.00002 | 0.000008 |
| Fast mode input | 0.000016 | not available | 0.000008 |
| Fast mode output | 0.00008 | not available | 0.00004 |
| Batch input | 0.000004 | 0.000005 | 0.000002 |
| Batch output | 0.00002 | 0.000025 | 0.00001 |
LLM pricing shown is based on hardcoded reference values and may not reflect current provider rates. Always verify against official pricing pages before making budget decisions.
What the sources say
Each rule in the calculator comes from one of these passages. The text below restates each passage in plain words and is not an exact quote. All were read on 2026-11-17.
Claude Fable 5.2: base input 8 dollars / MTok, 5m cache writes 10 dollars / MTok, 1h cache writes 16 dollars / MTok, cache hits and refreshes 0.20 dollars / MTok, output 40 dollars / MTok. MTok is Anthropic's unit for one million tokens.Anthropic pricing page, Model pricing table, Fable 5.2 row written out with its column names, platform.claude.com/docs/en/about-claude/pricing
Fable 5.2 fast mode: input 16 dollars / MTok, output 80 dollars / MTok. Up to 2.5 times higher output tokens per second. Prompt caching multipliers and data residency multipliers apply on top of fast mode pricing. Fast mode is not available with the Batch API.Anthropic pricing page, Fast mode pricing section, with the table row written out with its column names
claude-fable-5-2: batch input 4 dollars / MTok, batch output 20 dollars / MTok. All usage is charged at 50 percent of the standard API prices.Anthropic pricing page, Batch processing table, and the Anthropic batch processing guide
A 5 minute cache write costs 1.25 times the base input price and a 1 hour cache write costs 2 times the base input price. Cache hits and refreshes on Claude Fable 5.2 and Claude Mythos 5.2 (limited availability) cost 0.025 times the base input price. These multipliers stack with the Batch API discount and data residency.Anthropic pricing page, Prompt caching section
US-only inference via inference_geo applies a 1.1 times multiplier on all token categories, including cache writes and cache reads, on Claude 4.6 and later models.Anthropic data residency documentation, platform.claude.com/docs/en/manage-claude/data-residency
Claude Fable 5.2 uses adaptive thinking that is always on. A request that sets thinking to enabled with budget_tokens, or to disabled, returns a 400 error. The effort parameter (low, medium, high, xhigh, max) controls thinking depth, with a default of high on the API. Anthropic notes most of the capability gain is already present at high. Thinking tokens are billed as output tokens when they are generated, and thinking blocks of earlier turns that stay in context are billed as input tokens in later requests.What's new in Claude Fable 5.2, the effort documentation and the context windows documentation, platform.claude.com
Claude Fable 5.2 uses the tokenizer introduced with Claude Opus 4.7. This newer tokenizer produces approximately 30% more tokens for the same text than the tokenizer of earlier models, and the exact increase depends on the content. Fable 5.2 and Claude Mythos 5.2 (limited availability) require 30 day data retention and are not available under zero data retention unless expressly authorized.Anthropic pricing page (tokenizer note) and What's new in Claude Fable 5.2 (data retention), platform.claude.com
How effort level drives the bill
Adaptive thinking cannot be turned off on Fable 5.2, so thinking tokens are billed as output at 40 dollars per MTok. The effort parameter controls thinking depth. The API default is high, and Anthropic says most of the gain in capability is already present at high. Anthropic suggests xhigh for the hardest long-running coding and agentic work where you measured a gain, and max for the rare task where time and tokens do not matter. Claude Opus 5.5 defaults to medium, so a request that omits effort runs one level higher on Fable 5.2. Output at 40 dollars per MTok costs 5 times the input price of 8 dollars per MTok. No Anthropic page gives a number for the thinking tokens of each effort level, so measure your own output counts.
Questions
How much does Claude Fable 5.2 cost per million tokens?
8 dollars per MTok for input and 40 dollars per MTok for output on the standard synchronous API. 5 minute cache writes cost 10 dollars per MTok, 1 hour cache writes 16 dollars per MTok, and cache reads 0.20 dollars per MTok.
Are thinking tokens billed separately on Fable 5.2?
No. Adaptive thinking is always on and its tokens are billed as regular output tokens at 40 dollars per MTok.
Is fast mode worth 2 times the price?
Only if output latency is the bottleneck. Fast mode delivers up to 2.5 times higher output tokens per second for 2 times the standard input and output price, 16 and 80 dollars per MTok instead of 8 and 40. It is unavailable with the Batch API. Caching and the residency multiplier still apply.
Does the 1.1 residency multiplier apply to Fable 5.2?
Yes. US-only inference via inference_geo multiplies all token categories by 1.1 on Claude 4.6 and later models, which includes Fable 5.2. Fable 5.2 also requires 30 day data retention and is not offered under zero data retention without express authorization.
How do cache hits compare with other models?
Fable 5.2 and Mythos 5.2 (limited availability) use a 0.025 cache read multiplier, which is 0.20 dollars per MTok. Opus 5.5 and Sonnet 5.5 use 0.05, and all other models use 0.1. A Fable 5.2 cache read costs 0.20 dollars per MTok, the same as an Opus 5.5 cache read, although the base input price is 8 dollars per MTok on Fable 5.2 and 4 dollars per MTok on Opus 5.5.
What is the cheapest way to run Fable 5.2 at volume?
Combine the Batch API (4 and 20 dollars per MTok) with prompt caching for shared prefixes (reads at 0.20 dollars per MTok). The two stack. Fast mode and Batch are mutually exclusive.
Are these prices live?
No. They are reference rates, verified 2026-11-17. Check the source above before you set a budget.
Related tools
This page does not model retries or failed calls. The other tools cover cache savings and batch savings on other Claude models.
ClaudFlow is a free, privacy-first suite of AI workflow tools. No accounts. Tool inputs stay in your browser. Page and product-interest analytics is optional and requires consent. Part of the Zovo Tools network by Michael Lip.