Claude Haiku 5.5 pricing: what it costs and the fine print in the API

Claude Haiku 5.5 pricing: Anthropic launched it on October 7, 2026 and, in the API, it costs $0.10 input and $0.50 output per million tokens for prompts up to 100,000 tokens. Haiku 4.5 cost $1 and $5. That is a tenfold drop in list price, but with fine print: past 100,000 tokens the price goes up, and the new tokenizer counts more tokens for the same text.

What Claude Haiku 5.5 costs in the API

These are the prices per million tokens, according to Anthropic, for prompts up to 100,000 tokens:

• Input: $0.10 (Haiku 4.5: $1).
• Output: $0.50 (Haiku 4.5: $5).
• Cache reads: $0.01 per million tokens.

These are API prices, not app prices. If you use Claude through the app, this table does not change what you pay.

What it can do at that price

Haiku 5.5 has a 1-million-token context window and a maximum output of 128,000 tokens. That means it can read very long documents and write long answers. On top of that, cache reads at $0.01 per million make repeated prompts cheaper, like fixed instructions or documents you consult again and again.

The fine print on the price

A cheap model is only cheap if you know when it stops being cheap. There are two warnings:

1. The 100,000-token threshold. If the prompt goes past 100,000 tokens, the price rises to $0.50 input and $2.50 output per million: five times more.
2. The new tokenizer. It counts more tokens for the same text, so the real savings per task are smaller than ten times.

The practical takeaway: do not multiply your expected savings by ten. Measure how many tokens a typical task of yours uses with the new tokenizer and compare cost per task, not just price per million. For that logic, see how to compare models in Sonnet 5.5 vs Opus 5.5: cost per task.

How to apply it in your business?

When each call costs cents, counting call by call stops making sense and you can leave the model working on repetitive tasks all day. To start:

1. Pick a repetitive task with short prompts, under 100,000 tokens.
2. Reuse fixed instructions to take advantage of cache reads at $0.01 per million.
3. Measure real token usage with the new tokenizer.
4. Work out cost per task and compare it with your current model.
5. Watch long prompts: past 100,000 tokens, the price rises five times.

Frequently asked questions

How much does Claude Haiku 5.5 cost?

$0.10 input and $0.50 output per million tokens in the API, for prompts up to 100,000 tokens.

Is it ten times cheaper than Haiku 4.5?

In list price, yes: Haiku 4.5 cost $1 and $5. But the new tokenizer counts more tokens for the same text, so the savings per task are smaller than ten times.

What happens if my prompt goes past 100,000 tokens?

The price rises to $0.50 input and $2.50 output per million tokens.

Do these prices apply to the Claude app?

No. They are API prices, not app prices.

Conclusion

Claude Haiku 5.5 pricing is a good starting point for automating repetitive tasks, but read the fine print: the 100,000-token threshold and the new tokenizer. Confirm the figures on the Claude platform pricing page before budgeting.