Claude API pricing: what each model costs per request

Updated · 5 min read

The Claude API is billed per token, with separate prices for the text you send (input) and the text Claude writes (output). On 2026-11-18 Anthropic's prices per million tokens were $0.10 / $0.50 for Claude Haiku 5.5, $2 / $10 for Sonnet 5.5, $4 / $20 for Opus 5.5 and $8 / $40 for Fable 5.2. A typical chat turn on Sonnet costs about a cent. Prompt caching and the Batch API cut that further.

Current Claude API prices

Prices are per 1M tokens, standard tier, from Anthropic's pricing page on 2026-11-18:

Model Input Output Cache write (5 min) Cache write (1 hour) Cache hit Batch input / output
Claude Fable 5.2 $8 $40 $10 $16 $0.20 $4 / $20
Claude Opus 5.5 $4 $20 $5 $8 $0.20 $2 / $10
Claude Sonnet 5.5 $2 $10 $2.50 $4 $0.10 $1 / $5
Claude Haiku 5.5, prompts up to 100K tokens $0.10 $0.50 $0.125 $0.20 $0.01 $0.05 / $0.25
Claude Haiku 5.5, prompts over 100K tokens $0.50 $2.50 $0.625 $1 $0.05 $0.25 / $1.25

The other current models keep their full 1M-token context at the standard price: Anthropic says a 900K-token request is billed at the same per-token rate as a 9K-token one. Only Haiku 5.5 charges more for long prompts.

Cost calculator

Pick a preset or enter your own numbers. The cheapest model for that workload is at the top.

ModelPer requestPer month
Claude Haiku 5.5claude-haiku-5-5$0.10 in / $0.50 out per 1M$0.00045$0.67
Claude Sonnet 5.5claude-sonnet-5-5$2 in / $10 out per 1M$0.0090$13.50
Claude Opus 5.5claude-opus-5-5$4 in / $20 out per 1M$0.02$27.00
Claude Fable 5.2claude-fable-5-2$8 in / $40 out per 1M$0.04$54.00

USD list prices per 1M tokens, standard tier, checked on on each provider's pricing page. A month is 30 days. Taxes, batch discounts, tool fees (web search) and cache-write charges are left out.

Long prompts cost more (input / output): Claude Haiku 5.5 $0.50 / $2.50 for prompts over 100K tokens. The calculator switches to these rates when your input reaches them (marked *).

What a request costs in practice

  • A chat message is about 2,000 tokens in (your message plus the earlier conversation) and 500 out. On Sonnet 5.5 that is $0.004 + $0.005 = $0.009. On Haiku 5.5 it is $0.00045. $5 of credit covers about 550 Sonnet turns or about 11,000 Haiku turns.
  • A coding agent resends a large context on every step, often 40,000 tokens or more, but most of it comes from the cache. With 80% cache hits and 1,500 tokens of output, a Sonnet 5.5 step costs about $0.034 (8,000 fresh tokens $0.016, 32,000 cached tokens $0.0032, output $0.015). Three hundred steps a day comes to roughly $300 a month, before cache-write charges.
  • Summarizing a long document, 60,000 tokens in and 1,000 out, costs about $0.13 on Sonnet 5.5 and $0.0065 on Haiku 5.5.

These numbers use list prices and leave out cache writes, which cost 1.25 times the input price for the 5-minute cache and 2 times for the 1-hour cache.

How to pay less

  • Use prompt caching. A cache hit costs a twentieth of the normal input price on Sonnet and Opus 5.5 ($0.10 against $2 on Sonnet). Long system prompts, documents and conversation history are the parts worth caching. Many coding agents do this for you.
  • Use the Batch API for work that can wait. Anthropic processes batches asynchronously and charges 50% less for input and output. Batch and caching discounts can be combined.
  • Pick the smallest model that does the job. Haiku 5.5 costs a twentieth of Sonnet 5.5. It is fine for classification, extraction, routing and short replies.
  • Keep Haiku prompts under 100K tokens if you can. Above that, the whole request is billed at five times the rate.
  • Set a monthly spend limit under Spend limits in the Claude Console, so a runaway script can't drain your balance. You buy credits on the Billing page.

Charges on top of tokens

  • Web search: $10 per 1,000 searches, plus the tokens of the search results.
  • Web fetch: no extra charge, only the tokens of the fetched page.
  • Code execution: free when the request also includes the current web search or web fetch tool; otherwise billed by execution time.
  • Fast mode (up to 2.5 times faster output): $16 / $80 per million tokens on Fable 5.2 and $8 / $40 on Opus 5.5. It doesn't work with the Batch API.
  • US-only processing: for Claude 4.6 and later models, setting inference_geo to "us" multiplies prices by 1.1. The default global routing uses standard prices.

On Amazon Bedrock and Google Cloud (Google-managed), the cloud provider sets the price and bills you; check their pricing pages.

Older Claude models

You can still call several earlier models. Their prices per 1M tokens on 2026-11-18 were:

Model Input Output Cache hit
Claude Mythos 5.2 $8 $40 $0.20
Claude Fable 5.1, Mythos 5.1 $10 $50 $0.25
Claude Fable 5, Mythos 5 $10 $50 $1
Claude Opus 5, Opus 4.8, 4.7, 4.6, 4.5 $5 $25 $0.50
Claude Opus 4.1, Opus 4 $15 $75 $1.50
Claude Sonnet 5 $2 $10 $0.20
Claude Sonnet 4.6, 4.5, 4 $3 $15 $0.30
Claude Haiku 4.5 $1 $5 $0.10

Newer models are often cheaper than the ones they replace: Opus 5.5 costs less than Opus 5, and Haiku 5.5 a tenth of Haiku 4.5.

Tokens are not the same across providers

Anthropic says Claude 4.7 and later models use a tokenizer that produces about 30% more tokens for the same text than earlier Claude models. When you compare Claude with GPT, Gemini or DeepSeek on price per million tokens, the real cost per word can differ from what the table suggests. Our LLM API price comparison puts all major providers in one calculator.

API key or subscription?

A Claude Pro or Max plan is a flat monthly price with usage limits. The API bills every token with no monthly fee. If you use Claude now and then, or through several apps, the API is usually cheaper. If you run a coding agent for hours every day, a plan usually is; see using Claude Code with an API key for how to switch between the two. To create a key, follow how to get a Claude API key. For credits that come with a plan and other low-cost options, see free Claude API options.

Apps where you can use your Claude API key

  • Infinity

    Free · macOS

    Visit

    Free Mac AI chat app with git-style branching: fork any part of an answer into its own thread, ask up to four models at once, and use your own keys or local models.

    Works with:OpenAIAnthropicGoogle GeminixAI+7 more

  • Caotang

    Free · Web · Open source

    Visit

    Free, local-first browser writing studio for novels and scripts, with an AI agent that runs on your own Anthropic, OpenAI, Google, OpenRouter or local model key.

    Works with:AnthropicOpenAIGoogle GeminixAI+5 more

  • Laya

    Free · macOS, Windows, Linux · Open source

    Visit

    Open-source, local-first desktop command center that turns Slack, Gmail, GitHub, Jira and calendar notifications into AI-drafted action cards, using local models or your own API keys.

    Works with:AnthropicOpenAIGoogle GeminiOllama+1 more

  • Shizue

    Free · Chrome extension · Open source

    Visit

    Free, open-source Chrome side panel for AI chat, page summaries, bilingual video captions and in-tab tasks, running on your ChatGPT plan or your own OpenAI, Claude, Gemini or OpenRouter key.

    Works with:OpenAIAnthropicGoogle GeminiOpenRouter+1 more

  • PockoBots

    Free (beta) · Android · Open source

    Visit

    Open-source AI coding agent and personal bots that run on your Android phone, powered by your own OpenAI, Anthropic, Gemini, Groq or OpenRouter key with no app subscription.

    Works with:OpenAIAnthropicGoogle GeminiGroq+1 more

  • Exody

    Free · macOS, iOS

    Visit

    Free autonomous coding agent for Mac that routes each step to Claude, GPT, Gemini, Grok or a local model using your own API keys, with no seat fee or markup.

    Works with:AnthropicOpenAIGoogle GeminixAI

  • MergeJudge

    Free AI reviews on public repos; bring your own key for private repos · Web, Self-hosted · Open source

    Visit

    AI pull request reviewer offered as a hosted GitHub App or an open-source GitHub Action, running on your own Anthropic, OpenAI or OpenAI-compatible key, local Ollama, or rules only.

    Works with:AnthropicOpenAIOllama

  • Conduit

    Free · Windows, macOS, Linux · Open source

    Visit

    Open-source desktop AI app for agents, research, writing and slides that runs on your own keys from 23 providers or local models.

    Works with:AnthropicOpenAIGoogle GeminixAI+12 more

See all 89 apps that work with Anthropic

Questions

How much does the Claude API cost?

It is billed per token. On 2026-11-18 Anthropic listed Claude Haiku 5.5 at $0.10 per million input tokens and $0.50 per million output tokens, Sonnet 5.5 at $2 / $10, Opus 5.5 at $4 / $20 and Fable 5.2 at $8 / $40. A typical chat turn on Sonnet costs about one cent.

Is the Claude API free?

No, but Anthropic says new users receive free credits to test the API. After that you buy credits in the Claude Console and usage is deducted per token. See our guide to free Claude API options for the other ways to pay less.

What is the cheapest Claude model?

Claude Haiku 5.5, at $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens. Longer prompts are billed at $0.50 / $2.50.

Do batch and prompt caching discounts stack?

Yes. Anthropic says the Batch API discount (50% off input and output) and prompt caching discounts can be combined.

Why does Claude count more tokens than other models for the same text?

Anthropic says Claude 4.7 and later models use a newer tokenizer that produces about 30% more tokens for the same text. Keep that in mind when you compare per-token prices with other providers.

Is Claude cheaper through the API or with a Pro or Max plan?

It depends on how much you use it. A subscription is a flat monthly price with usage limits; the API has no limits beyond rate limits but bills every token. Light or occasional use is usually cheaper on the API; daily heavy use of a coding agent is usually cheaper on a plan.

Sources

Checked against the providers' own documentation on 2026-11-18. Prices and limits change; the links below have the current figures.

Related guides