The Claude Haiku 5.5 price looks like a bargain on the launch page: $0.10 per million input tokens, down from $1 on Haiku 4.5. But the 90% cut doesn’t cover every request, and the model counts text differently from its predecessor. Anthropic released it on 7 October, and the fine print decides whether your bill actually falls.
Claude Haiku 5.5 price: what Anthropic is charging
For prompts up to 100,000 tokens, Haiku 5.5 costs $0.10 per million input tokens and $0.50 per million output tokens. Above that line the rates rise to $0.50 and $2.50. Haiku 4.5 charged a flat $1 and $5. All prices are in US dollars.
Anthropic says the model costs around 75% less to run than Haiku 4.5 on average. That is not the same number as the 90%. By the company’s own footnote, it reflects both price tiers, the fact that about 90% of Haiku 4.5 requests stayed under 100,000 tokens, and the extra tokens the new tokenizer produces.
The rest of the spec sheet: a 1M-token context window, up from 200,000, and up to 128,000 output tokens. The API ID is claude-haiku-5-5. Anthropic says it is live on its own platform, AWS, Google Cloud and Microsoft Azure.
The two catches behind the 90%
The first is the 100,000-token line. Claude’s other recent models charge one rate across the whole 1M window. Haiku 5.5 doesn’t. Its price depends on prompt length, and the input rate jumps fivefold past the threshold. Feed it long documents and your saving drops to 50%.
The second is the tokenizer. Anthropic’s documentation says the same text counts as roughly 30% more tokens on Haiku 5.5 than on Haiku 4.5. A lower rate on more tokens is still cheaper, just not by as much as the headline suggests.
Here is our own arithmetic for a workload of 1 billion input and 200 million output tokens a month, counted the Haiku 4.5 way. We assumed the 30% inflation applies to input and output alike.
What is confirmed, and what is not
Confirmed vs not confirmed, as of 8 October 2026
Confirmed by Anthropic
- Released 7 October 2026.
- $0.10 / $0.50 per million tokens up to 100K; $0.50 / $2.50 above.
- 1M-token context window and 128K max output.
- API ID claude-haiku-5-5, live on the Claude Platform, AWS, Google Cloud and Azure.
- The same text counts as roughly 30% more tokens than on Haiku 4.5.
- Sonnet 5.5 beats it on every benchmark Anthropic published. Those scores are Anthropic’s own.
Not confirmed
- The workload mix behind the “around 75% less” average.
- Speed in tokens per second. VentureBeat says the launch materials it reviewed gave no figure.
- A coding score of about 20% at default effort. That is VentureBeat’s reading of Anthropic’s chart.
- GPT-6 Luna matching the price and adding a surcharge above 272,000 input tokens. VentureBeat’s report; OpenAI’s page not checked.
- Whether your AWS account can use it. Anthropic’s Bedrock page says “See Access”.
What Indian developers should check first
Before you switch from Haiku 4.5
- Recount your prompts with the model set to claude-haiku-5-5.
- Split your traffic into prompts up to 100K tokens and prompts over 100K.
- Remove assistant prefill, non-default temperature, top_p and top_k, and budget_tokens thinking. They now return errors.
- Handle the refusal stop reason in your client.
- Pick an effort level and test it. The default is medium.
- On Bedrock, check model access and your region before you plan.
- If you rely on Priority Tier, plan capacity separately. Haiku 5.5 doesn’t support it.
- Compare cost per finished task, not only price per token.
Haiku 5.5 is a genuine price cut for short-prompt work. Whether your bill falls by 90%, 75% or closer to 35% depends on prompt length and token counts. Measure first, then switch.
Claude Haiku 5.5 FAQ
How much does Claude Haiku 5.5 cost?
For prompts up to 100,000 tokens, it costs $0.10 per million input tokens and $0.50 per million output tokens. Above 100,000 tokens, the rates are $0.50 and $2.50. Prices are in US dollars, per Anthropic’s pricing page.
When was Claude Haiku 5.5 released?
Anthropic released it on 7 October 2026. The API ID is claude-haiku-5-5, and Anthropic says it is available on its own platform, Amazon Web Services, Google Cloud and Microsoft Azure.
How much cheaper is Claude Haiku 5.5 than Haiku 4.5?
Anthropic prices it 90% lower for requests up to 100,000 tokens and 50% lower above that. Its own average estimate is around 75% less to run, which also counts the new tokenizer, where the same text becomes roughly 30% more tokens.
Sources
Sources: Anthropic announcement · pricing · what’s new · migration guide · Bedrock docs · VentureBeat. Last Updated: 8 October 2026.