Claude through the API

The Claude family from Anthropic is regarded as one of the strongest model lines for code and long-form writing. The current generation consists of Opus 4.1 (top model), Sonnet 4.5 (workhorse) and Haiku 4.5 (fast and cheap). Here are the rates and considerations for use from the EU.

Rates per million tokens

Claude Opus 4.1: $15 input / $75 output per million tokens. Claude Sonnet 4.5: $3 / $15. Claude Haiku 4.5: $1 / $5. Context window is 200k tokens for the whole family. Opus is thereby the most expensive model in the regular offering; Sonnet is the model most production tasks run on.

When Opus, when Sonnet or Haiku

Opus for the hardest tasks: complex refactors, agentic flows, deep analysis. Sonnet covers the broad work: code review, document generation, structured outputs. Haiku for high volumes: classification, extraction, chat completion. The price ratio Opus:Haiku is 15:1 on input — model choice is the biggest cost lever.

Using Claude from the EU

Anthropic contracts EU customers via an Irish entity and offers EU data processing. Invoicing happens in dollars via credit card. Through Gatewaybird (from launch) you call Claude via one OpenAI-compatible endpoint with a Dutch contract and euro invoice, alongside the other providers in the same offering.

Alternatives at a similar level

Sonnet competes on price/quality with GPT-5 and Gemini 2.5 Pro. For pure price power, the open models (Llama, Qwen, DeepSeek) are 10-50x cheaper — often suitable for part of the tasks. Compare in the price comparison.

Prijsvergelijker →OpenAI API vanuit de EU →

Frequently asked questions

What does the Claude API cost?

Opus 4.1: $15/$75 per million input/output tokens. Sonnet 4.5: $3/$15. Haiku 4.5: $1/$5. Prices from Anthropic, August 2026.

Which Claude model is best for code?

Claude Sonnet 4.5 is the workhorse for code; Opus 4.1 for the hardest tasks. For code review routines, Haiku is often sufficient already.

Is there a cheaper alternative of similar quality?

Gemini 2.5 Pro and GPT-5 sit at a similar quality level with different pricing. Open-weight models like DeepSeek V3.1 are 10x cheaper with part of the quality — suitable when it is not the heaviest task.

Can I use Claude with EU invoicing?

Through an EU intermediary like Gatewaybird: yes, after launch. Directly, Anthropic invoices in dollars.

Compare Claude rates

One email at launch. No spam.