Anthropic API pricing per million tokens

How much does the Claude API cost per million tokens?

Anthropic charges per million tokens, with separate rates for input, cache writes, cache hits and output, and each Claude model has its own rates. All prices on the Anthropic pricing page are in USD.

Model	Input	5-minute cache write	1-hour cache write	Cache hit	Output
Claude Fable 5.1	$10	$12.50	$20	$0.25	$50
Claude Opus 5.5	$4	$5	$8	$0.20	$20
Claude Opus 5	$5	$6.25	$10	$0.50	$25
Claude Sonnet 5.5	$2	$2.50	$4	$0.10	$10
Claude Sonnet 5	$2	$2.50	$4	$0.20	$10
Claude Sonnet 4.6	$3	$3.75	$6	$0.30	$15
Claude Haiku 5.5, prompts up to 100,000 tokens	$0.10	$0.125	$0.20	$0.01	$0.50
Claude Haiku 5.5, prompts over 100,000 tokens	$0.50	$0.625	$1	$0.05	$2.50
Claude Haiku 4.5	$1	$1.25	$2	$0.10	$5

A few notes from the same page.

• Claude Opus 4.5 through 4.8 are listed at the same rates as Claude Opus 5.
• Claude Mythos 5.1 is listed at the same rates as Claude Fable 5.1, with limited availability.
• Claude Sonnet 5’s $2 input and $10 output rates began as introductory pricing and are now the standard price.
• Claude 4.7 and later models, and Claude Mythos Preview, use a newer tokenizer that “produces approximately 30% more tokens for the same text.”

How does prompt caching change the price?

Cache reads cost a fraction of base input, while cache writes cost more than base input. The pricing page sets these multipliers against the base input rate.

Cache operation	Multiplier	How long the cache lasts
5-minute cache write	1.25x base input	5 minutes
1-hour cache write	2x base input	1 hour
Cache hit	0.1x base input on most models	Same as the write

Cache hits cost 0.05x on Claude Opus 5.5 and Claude Sonnet 5.5, and 0.025x on Claude Fable 5.1 and Claude Mythos 5.1. Per Anthropic, caching “pays off after one cache read for the 5-minute duration (1.25x write), or after two cache reads for the 1-hour duration (2x write).” These multipliers stack with the Batch discount and data residency pricing.

How much does the Message Batches API save?

The Batch API halves the price, with “a 50 percent discount on both input and output tokens” per the pricing page. The Message Batches guide says most batches finish in less than 1 hour, and requests that error, are canceled or expire are not billed.

Model	Batch input	Batch output
Claude Fable 5.1	$5	$25
Claude Opus 5.5	$2	$10
Claude Opus 5	$2.50	$12.50
Claude Sonnet 5.5	$1	$5
Claude Sonnet 5	$1	$5
Claude Sonnet 4.6	$1.50	$7.50
Claude Haiku 5.5, prompts up to 100,000 tokens	$0.05	$0.25
Claude Haiku 5.5, prompts over 100,000 tokens	$0.25	$1.25
Claude Haiku 4.5	$0.50	$2.50

What does fast mode cost?

Fast mode is a premium rate for faster output on three Opus models. Fast mode for Claude Opus 5 and Claude Opus 4.8 is in research preview, per the pricing page.

Model	Input	Output
Claude Opus 5.5	$8	$40
Claude Opus 5 and Claude Opus 4.8	$10	$50

Prompt caching and data residency multipliers apply on top of fast mode rates. Fast mode is not available with the Batch API.

Are there other multipliers?

Yes, for US-only inference. Per the pricing page, setting inference_geo to "us" on Claude 4.6 and later models applies a 1.1x multiplier to all token prices, including cache writes and reads. Global routing, the default, uses standard pricing.

Claude on Amazon Bedrock and Google Cloud (Google managed) is priced by those providers. On those platforms, regional and multi-region endpoints carry a 10 percent premium over global endpoints for Claude 4.5 models and later.

How do you pay for Claude API usage?

Most organizations pay in advance. Anthropic’s billing article says most organizations use prepaid usage credits, which expire one year from purchase. Failed requests aren’t charged. The full picture, with invoices, tiers and spend limits, is in how OpenAI and Anthropic API billing works.

To turn these rates into a monthly figure, try the AI API cost calculator. For how Claude charges show up on invoices, read Anthropic API billing explained on the ernd blog.