Claude Fable 5.2 Pricing

Claude Fable 5.2 price per million tokens

Anthropic's published list prices, USD per million tokens. The right-hand column is our own division of each rate by the $8 base input price — the same multipliers Anthropic publishes for caching, batching and fast mode.

Sources: Anthropic's pricing documentation and the claude.com pricing page, both read November 11, 2026. Batch bills 50% off input and output for asynchronous work. Fast mode bills exactly 2x for up to 2.5x faster output and is not available with the Batch API. Prompt-caching and data-residency multipliers stack on top of the batch and fast-mode rates.

Every Claude and rival model rate, side by sideFast mode pricing in detail

Fable 5.2 vs Fable 5.1, Fable 5 and Opus 5.5

The same line items across four Anthropic models. Fable 5.2 replaced Fable 5.1 as the Fable-tier flagship at 20% lower input and output prices; Opus 5.5 is the cheaper model Anthropic tells you to start on.

Figures are list prices per million tokens from this site's model registry: the Fable 5.2 row was verified against Anthropic's pages on November 11, 2026, while the Fable 5.1, Fable 5 and Opus 5.5 rows keep their own earlier verification dates. Fast mode is 2x each model's base rate and Anthropic offers it for Fable 5.2 and Opus 5.5 only. On retention, the Fable 5.2 model card states that 30-day retention is required and zero data retention is not available unless expressly authorised.

Fable 5.2 vs Opus 5.5, line by lineFable 5.1 vs Fable 5

What changed from Fable 5.1, and how Fable 5.2 sits against Opus 5.5

Both deltas are simple division on published list prices, not a measurement of what any workload costs.

Input is 20% lower than Fable 5.1 and output 20% lower, which is where Anthropic's "$10/$50 to $8/$40" framing comes from. The cache-hit multiplier changed too: Fable 5.2 bills 0.025x base input instead of the 0.1x standard rate. Cache writes stay on Anthropic's published multiples — 1.25x for 5 minutes and 2x for 1 hour — so until the cache is read, the write costs more than it saves. Against Opus 5.5 the sticker price is 2xx on both input and output.

Input / MTok

$8.00

Output / MTok

$40.00

Input vs Fable 5.1

−20%

Price vs Opus 5.5

2xx

Percentages and ratios are our arithmetic on registry list prices. Anthropic has not published a dollar cost-per-task comparison for Fable 5.2; its launch material makes only a qualitative claim — about half the cost per task versus Opus 5.5 at max effort on FrontierCode 2.0 — which is a vendor statement about one benchmark and is not converted into a number here.

What stacks on top of the $8/$40 rate

Five factors change the final bill. Four of them are published rules rather than extra prices; the figures below are what those rules produce at Fable 5.2's rates.

• 1M context at standard pricing

  Fable 5.2 serves its full 1M-token context window at the same per-token rate — Anthropic states that Claude 4.6 and later models (except Haiku 5.5) include the full window at standard pricing, so a 900K-token request bills at the same rate as a 9K-token one. Max output is 128K tokens.

• 30-day retention is mandatory

  Fable 5.2 is a Covered Model: Anthropic's model card states that it requires 30-day data retention and is not available under zero data retention unless Anthropic expressly authorises it. That is a compliance constraint rather than a line on the invoice, but it decides whether some workloads can use the model at all.

• Fast mode: 2x, and no Batch

  Anthropic lists fast mode for Fable 5.2 at $16 input / $80 output per MTok — exactly double — for up to 2.5x faster output. It is not available with the Batch API, so a job is either cheap and asynchronous or fast and expensive. Prompt-caching and data-residency multipliers apply on top of the fast rate.

• A cache write costs more than it saves, until it is read

  A 5-minute cache write bills 1.25x base input ($10/MTok) and a 1-hour write 2x ($16/MTok); reads bill 0.025x ($0.20/MTok). Anthropic's own rule of thumb is that caching pays off after one read at the 5-minute duration and after two reads at the 1-hour duration. Cache multipliers stack with the Batch discount and with fast mode.

• US-only inference adds 10%

  Setting inference_geo to "us" applies a 1.1x multiplier to every token category — input, output, cache writes and cache reads. Global routing, the default, bills at standard pricing. Partner-operated platforms — Amazon Bedrock and Google Cloud's Google-managed offering — set their own regional prices.

Rules quoted from Anthropic's pricing documentation, read November 11, 2026. The dollar figures are the registry rates multiplied by Anthropic's published multipliers.

Fast mode pricing in detailPrompt caching pricing, TTL by TTL

Worked example: what a month looks like

A hypothetical volume — 10 million input and 2 million output tokens a month — priced at list rates. It is arithmetic to show the shape of the bill, not a measured cost per task and not a recommendation for your workload.

Hypothetical workload: 10M input tokens and 2M output tokens per month, no cache hits, no batch discount, global routing.

Fable 5.2

$160.00

Fable 5.1

$200.00

Saved vs Fable 5.1

$40.00

Opus 5.5

$80.00

The figures are list-price arithmetic on registry rates, not a quote. A real bill depends on cache hit rate, batch eligibility, effort level and tokenizer drift — Anthropic notes that Claude 4.7 and later use a tokenizer producing roughly 30% more tokens for the same text, which moves the bill without moving the per-token price.

Model your own monthly token billEvery Claude and rival model rate, side by side

Keep reading