Claude Fable 5, Fable 5.1 and Opus 5.5 API pricing: cost per million tokens

Claude Fable 5.1, Claude Fable 5 and Claude Opus 5.5 are billed per token on SeedRouter, with separate rates for input, output, and the two prompt-cache operations. Claude Opus 5.5 is the cheapest of the three per token. Claude Fable 5.1 and Claude Fable 5 cost the same for input and output, but Fable 5.1 reads cached prompts at a quarter of Fable 5's cache-read rate. There is no subscription, and a request that fails is not charged.

The tables below are read from the rate table that bills your account, so they show what a request costs today.

What do Claude Fable 5.1, Fable 5 and Opus 5.5 cost per million tokens?

Each model has one table. Input and output are the base rates. "Output" includes the model's thinking, because thinking tokens are output tokens. The cache columns apply only when you use prompt caching. If a rate cannot be loaded, the table says so instead of showing zero; treat it as unknown, never as free.

For reference, this is how Anthropic's list prices on its own API are structured, from the Claude pricing page as of September 27, 2026. Each line is a multiple of that model's own input rate:

On that list, Claude Fable 5.1 and Claude Fable 5 share one input rate, and Claude Opus 5.5's input rate is 40% of it.

How is one Claude request billed?

Every response carries a usage object, and the charge is computed from it, not from your max_tokens:

cost = ( input_tokens × input rate
       + output_tokens × output rate
       + cache_read_input_tokens × cache-read rate
       + 5-minute cache writes × 5-minute write rate
       + 1-hour cache writes × 1-hour write rate ) / 1,000,000

In the Messages format, the two kinds of cache writes are reported separately in usage.cache_creation.ephemeral_5m_input_tokens and usage.cache_creation.ephemeral_1h_input_tokens.

max_tokens is only a ceiling. A request with max_tokens: 32000 that writes 900 tokens pays for 900.

Which Claude model is cheapest for my workload?

The same SeedRouter key calls all three, so switching is a change to model. The Claude Fable 5.1 vs Fable 5 vs Opus 5.5 guide covers what else changes between them.

How much does prompt caching save?

Caching pays when the same long prefix, such as a system prompt, a codebase or a document, is sent again and again. The first request writes the prefix to the cache, which costs more than plain input. Later requests read it back at a small fraction of the input rate.

On Anthropic's list, a 5-minute write costs 1.25 times the input rate and a 1-hour write costs 2 times. A cache read costs 2.5% of the input rate on Claude Fable 5.1, 5% on Claude Opus 5.5 and 10% on Claude Fable 5. The cache columns in the tables above are SeedRouter's own rates for the same five lines.

At those ratios a 5-minute write pays for itself the first time the prefix is read back, and a 1-hour write from the second read. Use the 1-hour write when requests are more than five minutes apart. To turn caching on, add cache_control to the request as described in the Claude Opus 5.5 API reference.

Does the effort setting change the price?

Not the rate, but the bill. Effort (output_config.effort: low, medium, high, xhigh or max) decides how much the model thinks, and thinking is billed as output. The same question at low usually costs less than at max. Claude Fable 5.1 and Claude Fable 5 default to high; Claude Opus 5.5 defaults to medium.

Thinking cannot be switched off on these three models, only turned down. If a task is simple, lower effort is the most direct way to cut cost.

Is a Claude Pro or Max subscription the same as API access?

No. Anthropic's support article is explicit: "A paid Claude subscription enhances your chat experience but doesn't include access to the Claude API or Console" (source). Pro and Max cover the Claude apps. API calls are billed per token, separately.

On SeedRouter there is only the per-token side. You top up once on the billing page, the credits never expire, and each finished request is deducted from that balance. The same balance pays for every model in the catalogue.

Frequently asked questions

How much do 1 million Claude tokens cost?

It depends on the model and on whether the tokens are input or output. Output costs five times input on all three models. The live tables above give the rate for each, and the Claude Opus 5.5 row is the lowest.

Is Claude Opus 5.5 cheaper than Claude Fable 5.1?

Yes, on every line of the table. On Anthropic's list, Claude Opus 5.5's input and output rates are 40% of Claude Fable 5.1's, and the SeedRouter rates above keep that order.

Is Claude Opus 5.5 cheaper than Claude Opus 5?

On Anthropic's list, yes: its input and output rates are 20% below Claude Opus 5's, and cache reads are 60% below. SeedRouter does not sell Claude Opus 5.

Does a long prompt cost more per token?

No. Anthropic bills the full 1M-token context window of these models at the standard rate: its pricing page says "a 900k-token request is billed at the same per-token rate as a 9k-token request." A long prompt costs more only because it has more tokens.

Are Batch API discounts or fast mode available?

Not on SeedRouter. Requests run through the standard Messages, Chat Completions and Responses endpoints and are billed at the rates in the tables above.

Where can I see what a request cost?

In usage history, request by request. The API response itself reports token counts in usage but no dollar amount.

Price it before you send it

Start on Claude Opus 5.5, move a task to Claude Fable 5.1 only when it needs deeper reasoning, and cache any prefix you send more than once. Check the live tables above before a large job, top up on the billing page, and see the Claude Fable 5.1, Claude Fable 5 and Claude Opus 5.5 pages to try each model in the browser.