Claude API Pricing

Claude API prices are per million tokens, from $0.10 input and $0.50 output on Claude Haiku 5.5 to $10 and $50 on Claude Fable 5.1, with Opus 5.5 at $4 and $20 and Sonnet 5.5 at $2 and $10, as of Nov 6, 2026. Prompt caching and the 50% Batch discount lower that.

Claude API prices on Nov 6, 2026 are, per million tokens of input and output: Claude Opus 5.5 $4 and $20, Claude Sonnet 5.5 $2 and $10, Claude Haiku 5.5 $0.10 and $0.50, and Claude Fable 5.1 $10 and $50. Output always costs five times input. Two discounts apply on top: Prompt Caching bills repeated input at a tenth of the input price or less, and the Message Batches API takes 50% off everything, and the two stack. Billing is in US dollars on actual monthly use, by credit card or, for enterprise accounts, invoice; Anthropic's docs name claude.com/pricing as the most current price list. This page covers the Claude API sold by Anthropic; Amazon Bedrock and Google's own Claude service set their own prices (Claude on Cloud Platforms).

Quick facts

Cheapest model	Claude Haiku 5.5, $0.10 in / $0.50 out (prompts up to 100,000 tokens)
Recommended default	Claude Opus 5.5, $4 in / $20 out
Most capable open model	Claude Fable 5.1, $10 in / $50 out
Cache read	0.1x input on most models; 0.05x on Opus 5.5 and Sonnet 5.5; 0.025x on Fable 5.1
Cache write	1.25x input (5 minutes) or 2x input (1 hour)
Batch discount	50% on input and output
1M-token context	Same per-token price at any length (Claude 4.6 and later, except Haiku 5.5)
Web search	$10 per 1,000 searches, plus tokens
Managed Agents runtime	$0.08 per session-hour, plus tokens

All values from Anthropic's Pricing page, read Nov 6, 2026. MTok means a million tokens.

Price per model

Model	Input	5-minute cache write	1-hour cache write	Cache read	Output
Claude Fable 5.1	$10	$12.50	$20	$0.25	$50
Claude Fable 5	$10	$12.50	$20	$1	$50
Claude Opus 5.5	$4	$5	$8	$0.20	$20
Claude Opus 5, 4.8, 4.7, 4.6, 4.5	$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, 4.5	$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

US dollars per million tokens, from Pricing, Anthropic docs, read Nov 6, 2026. Claude Mythos 5.1 and Mythos 5 cost the same as the Fable model of the same number but are sold only to verified organizations.

Three quirks are worth knowing. Opus 5.5 is cheaper than the Opus models before it ($4 against $5 input). Sonnet 5 was announced at launch with an introductory $2 and $10 through Aug 31, 2026 and a rise to $3 and $15 after; Anthropic's pricing page now says that rise will not occur. And Sonnet 5.5's cache reads were cut from $0.20 to $0.10 on Oct 7, 2026.

Tokens are not words

Prices are per token, and the tokenizer changed. Claude 4.7 and later models use a tokenizer that produces about 30% more tokens for the same text than Sonnet 4.6 and earlier, so a move from Sonnet 4.6 ($3) to Sonnet 5.5 ($2) saves less than the list prices suggest. On the current tokenizer, a million tokens is roughly 555,000 words. The free POST /v1/messages/count_tokens endpoint counts a request before you send it (API overview).

Long context, US-only inference and fast mode

Long context: Claude 4.6 and later models (except Haiku 5.5) bill the whole 1M-token window at the standard rate; in Anthropic's words, a 900k-token request costs the same per token as a 9k-token one. Haiku 5.5 is the exception: once a prompt passes 100,000 tokens (cache reads and writes included), the whole request pays the higher Haiku prices.
US-only inference: setting inference_geo: "us" on Claude 4.6 and later models costs 1.1x on every token category; the default global routing is the standard price.
Fast mode: speed: "fast" gives up to 2.5x faster output at a premium: $8 input and $40 output on Opus 5.5, and $10 and $50 on Opus 5 and Opus 4.8 (a research preview there; Opus 4.8's fast mode ends Nov 30, 2026). Fast mode does not work with batches.

Tool prices

Tool use is billed as ordinary tokens: the tool definitions, the tool_use and tool_result blocks, and a hidden tool system prompt of 286 tokens on Opus 5.5, Sonnet 5.5 and Haiku 5.5. Some server tools add a fee (Pricing):

Tool	Price
Web search	$10 per 1,000 searches, plus result tokens; failed searches are free
Web fetch	No fee beyond tokens (a 10 kB page is about 2,500 tokens)
Code execution	Free when web search or web fetch version `20260209` or later is in the request; otherwise 1,550 free hours per organization each month, then $0.05 per container-hour, 5-minute minimum
Computer Use toolset	About 4,500 extra input tokens per request, plus screenshots as images
Managed Agents	$0.08 per session-hour while running, plus tokens; no batch discount

Source: Pricing, Anthropic docs, read Nov 6, 2026.

Batch prices

The Message Batches API halves every price for work that can wait (most batches finish within an hour, all within 24). Per million tokens, input and output: Fable 5.1 $5 and $25, Opus 5.5 $2 and $10, Sonnet 5.5 $1 and $5, Haiku 5.5 $0.05 and $0.25 (up to 100,000-token prompts). Caching multipliers apply on top of the batch price.

A worked example

A support bot on Sonnet 5.5 sends 1,000 requests in an hour. Each carries the same 10,000-token system prompt plus 2,000 new tokens and gets a 500-token answer, with no gap longer than five minutes, so the cache never expires. At the Sonnet 5.5 prices:

	Without caching	With 5-minute caching
System prompt	10M tokens x $2 = $20.00	one write, 10,000 x $2.50/MTok = $0.03; 999 reads, 9.99M x $0.10 = $1.00
New input	2M x $2 = $4.00	2M x $2 = $4.00
Output	0.5M x $10 = $5.00	0.5M x $10 = $5.00
Total	$29.00	about $10.03

The same workload on Opus 5.5 costs $58.00 uncached. Sent as a batch without caching, the Sonnet run costs $14.50, though answers arrive later. Anthropic's own worked example for Managed Agents prices a one-hour Opus 5.5 session with 50,000 input and 15,000 output tokens at $0.58.

Free credits, plan credits and discounts

New Console accounts get a small amount of free credits; Anthropic does not state the amount. Claude Max, Ultra and Team subscribers receive monthly API credits: $100 (Max), $200 (Ultra), and $20 per Standard or $100 per Premium Team seat, pooled up to $500, with no rollover and no use in Claude Code. Volume discounts are negotiated case by case, and academic discounts "might be available" (Pricing). Monthly spend is also capped by usage tier, from $500 in the Start tier to $200,000 in Scale (Claude API Rate Limits).

Since Sep 24, 2026, a refusal before any output is billed when its category is bio, frontier_llm or reasoning_extraction; other refusals before output stay free.

Frequently asked questions

How much does the Claude API cost per token? As of Nov 6, 2026 the Claude API charges per million tokens: Opus 5.5 $4 input and $20 output, Sonnet 5.5 $2 and $10, Haiku 5.5 $0.10 and $0.50, and Fable 5.1 $10 and $50 (Pricing).

What is the cheapest Claude API model? Claude Haiku 5.5 is the cheapest Claude model, at $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens, and $0.05 and $0.25 through the Batch API (Pricing).

Does Claude charge more for long prompts? Not on Claude 4.6 and later models: Opus 5.5, Sonnet 5.5 and Fable 5.1 bill the whole 1M-token context at the standard rate. Claude Haiku 5.5 is the exception, with higher prices for prompts over 100,000 tokens.

Is the Claude API cheaper than a Claude subscription? It depends on volume: the Claude API bills every token, while Pro and Max are flat monthly prices with usage limits. Light or automated use is often cheaper on the API, and heavy interactive use, especially in Claude Code, is often cheaper on a plan (Claude Plans and Pricing).

How do I lower my Claude API bill? Cache repeated context with prompt caching (cache reads cost 0.1x input or less), send non-urgent work through the Message Batches API at 50% off, and use the smallest Claude model that does the job well.