Claude API Pricing (2026): Token Rates & Cost Per Token

Claude API pricing is pay-as-you-go: Anthropic charges per million tokens (MTok), with one rate for the tokens you send (input) and a higher rate for the tokens Claude writes back (output). As of November 10, 2026, the default general model, Claude Sonnet 5.5, costs $2 per million input tokens and $10 per million output tokens. The fastest model, Haiku 5.5, starts at $0.10 / $0.50, and the newest flagship, Fable 5.2, is $8 / $40.

This guide covers the full Claude API price list, the Claude Opus API price, what prompt caching and the Batch API do to your bill, and a simple method to estimate Claude token cost for your own workload. Every number comes from Anthropic's official Claude API pricing docs and models overview, both checked on November 10, 2026. Prices are in US dollars and change often, so confirm on those pages before you budget.

If you want to chat with Claude in its own apps rather than call it from code, you need a plan, not the API — see Claude Pro vs Max.

What Claude API pricing covers (and how it differs from Pro and Max)

The Claude API is for developers who call Claude models from their own app, script or agent. You create an account in the Claude Console, add credit, and are billed on actual monthly usage. Anthropic's docs say new users get a small amount of free credit to test the API; after that, every request is billed.

A Claude plan (Free, Pro, Max, Ultra, Team) is different. It pays for using Claude's own apps and Claude Code, with usage limits rather than token bills. The only overlap: Max includes $100 a month and Ultra $200 a month in API credits, and Team plans include up to $500 a month in pooled API credits, per Anthropic's pricing page.

So the question "is the Claude API cheaper than Pro?" has no single answer. A plan is a flat fee for a person using an app. The API is a metered bill for software. Compare them only against your real usage.

Claude API price list: official token rates at a glance

These are the current models in Anthropic's lineup, with standard (non-batch) rates. Cache read is the price for input tokens served from the prompt cache.

Anthropic's models page says that if you are unsure, start with Opus 5.5 for most workloads, and use Fable 5.2 for the most demanding reasoning. All four current models list a 1M-token context window and 128K max output. Claude Mythos 5.2 is listed at the same rates as Fable 5.2 but with limited availability — only for organizations verified through Anthropic's verification programs.

Older models still on the price list

Legacy models stay available at their own rates. A few examples from the same page:

Model	Input	Output
Claude Fable 5.1	$10	$50
Claude Opus 5, 4.8, 4.7, 4.6, 4.5	$5	$25
Claude Sonnet 5	$2	$10
Claude Sonnet 4.6 and 4.5	$3	$15
Claude Haiku 4.5	$1	$5

If you are still on an older model, check whether the newer one is cheaper: Opus 5.5 ($4 / $20) undercuts Opus 4.5 to Opus 5 ($5 / $25), and Haiku 5.5 is a tenth of Haiku 4.5's input price for prompts under 100K tokens.

Claude Opus API price

The Claude Opus API price for the current model, Opus 5.5, is $4 per million input tokens and $20 per million output tokens (November 10, 2026). Other Opus rates on the official sheet:

• Cache writes: $5/MTok (5-minute cache) or $8/MTok (1-hour cache)
• Cache reads: $0.20/MTok — 5% of the base input price on this model
• Batch API: $2 input / $10 output
• Fast mode: $8 input / $40 output, for significantly faster output at a premium

Fast mode is available on the Claude API for Fable 5.2 ($16 / $80), Opus 5.5 ($8 / $40) and Opus 5 / Opus 4.8 ($10 / $50, in research preview). It cannot be combined with the Batch API.

Sonnet and Haiku token costs

Claude Sonnet 5.5 is the mid-priced workhorse: $2 input / $10 output, with cache reads at $0.10 and batch at $1 / $5. For many production apps — chat, drafting, tool use — Sonnet is where the cost-to-quality trade-off usually starts.

Claude Haiku 5.5 is priced by prompt length. Requests with prompts up to 100,000 tokens pay $0.10 / $0.50; requests over that pay $0.50 / $2.50. Anthropic counts every input token in the request toward that threshold, including cache reads and writes. If you route, classify or extract at high volume with short prompts, Haiku is the cheapest current option on the sheet.

How to estimate Claude cost per token for a real workload

Anthropic does not publish an "average" API bill, and neither will we — usage varies too much. Instead, use this method:

1. Count tokens for a typical request. Use Anthropic's token counting endpoint with the exact model ID. The docs note that Claude 4.7 and later models use a newer tokenizer that produces about 30% more tokens for the same text than older models, so count on the model you will actually use.
2. Split input and output. Input is your system prompt, history, documents and tool definitions. Output is what Claude writes, including any thinking it does.
3. Multiply by the rates. Cost per request = (input tokens ÷ 1,000,000 × input rate) + (output tokens ÷ 1,000,000 × output rate).
4. Multiply by volume. Requests per day × 30 gives a monthly estimate.
5. Add extras. Web search, code execution or Managed Agents runtime if you use them (see below).

Worked example (arithmetic only, not a measured workload): a request with 3,000 input tokens and 800 output tokens on Sonnet 5.5 costs (3,000 ÷ 1,000,000 × $2) + (800 ÷ 1,000,000 × $10) = $0.006 + $0.008 = $0.014. At 1,000 such requests a day for 30 days, that is about $420 a month before caching or batching. The same request on Haiku 5.5 is $0.0003 + $0.0004 = $0.0007, or about $21 a month at the same volume.

Anthropic's own docs include a similar example for a Managed Agents session on Opus 5.5: 50,000 input and 15,000 output tokens plus one hour of session runtime comes to $0.58.

To test a prompt without writing code, paste it into our free AI token calculator.

Other official Claude Platform fees

Tokens are most of the bill, but a few features add their own charges (official docs, November 10, 2026):

Feature	Price
Web search	$10 per 1,000 searches, plus tokens for the results
Web fetch	No extra charge; you pay for the fetched tokens
Code execution	1,550 free hours per organization per month, then $0.05 per hour per container; free when used with the newer web search or web fetch tools
Managed Agents	$0.08 per session-hour of running time, plus tokens
US-only inference (inference_geo: "us")	1.1x multiplier on all token prices for Claude 4.6 and later models

Tool use also adds a small fixed system prompt: for example 286 tokens on Opus 5.5, Sonnet 5.5 and Haiku 5.5 when tools are present, per the docs.

Ways to cut Claude API costs

These are the levers Anthropic documents. We do not quote a percentage saving for your app, because it depends on your traffic.

• Use the Batch API for non-urgent work. Batch requests are 50% off both input and output tokens. Sonnet 5.5 drops to $1 / $5.
• Turn on prompt caching. A 5-minute cache write costs 1.25x base input and a 1-hour write costs 2x. Reads cost 0.1x base input on most models, 0.05x on Opus 5.5 and Sonnet 5.5, and 0.025x on Fable 5.2 and 5.1. Anthropic notes caching pays off after one read for the 5-minute cache. Batch and caching discounts can be combined.
• Pick the smallest model that passes your tests. Anthropic's own guidance: Haiku for simple tasks, Sonnet for most production workloads, Opus for complex reasoning.
• Keep prompts short. Long system prompts and full chat histories are input tokens on every call. Trim what the model does not need.
• Cap output length. Output costs five times input on every current model, so set max_tokens to what the task needs.
• Watch long prompts on Haiku. Crossing 100K tokens moves a whole request to the higher Haiku rate.
• Ask about volume pricing. Anthropic says volume discounts are negotiated case by case through sales.

Claude API vs a Claude plan vs Claude Code

• Chatting, writing, research in Claude's apps: a plan. See Claude Pro vs Max.
• Coding with Claude Code in your terminal or editor: included on paid plans. See Claude Code pricing.
• Calling Claude from your own product: the API, priced on this page.
• Comparing providers: see OpenAI API pricing and Gemini API pricing for the other official sheets.

For what each model is actually for, read Claude models explained. For choosing between assistants rather than APIs, see Claude vs ChatGPT and our ChatGPT alternatives guide.

FAQ