Anthropic released Claude Haiku 5.5 on October 7. It costs $0.10 per million input tokens and $0.50 per million output tokens for requests up to 100,000 tokens. It is the first Haiku model with effort controls. Anthropic also halved a Sonnet 5.5 price and added monthly API credits for some plans.
New to this? Read it in simple words
- Anthropic released Claude Haiku 5.5, its new small and cheap AI model.
- Shorter requests cost $0.10 per million input tokens, down from $1.
- An outside test ranks it second of 182 models in its class.
- It uses many tokens, so the real cost depends on the job.
- Token
- A small piece of text, often part of a word. AI models are priced per million tokens.
- Effort control
- A setting that tells the model how long to think before it answers.
- Cache read
- Reusing a stored part of a prompt, which costs less than sending it again.
The new prices
For requests up to 100,000 tokens, Haiku 5.5 costs $0.10 per million input tokens and $0.50 per million output tokens. Haiku 4.5 cost $1 and $5, so these prices are 90% lower.
Longer requests cost $0.50 for input and $2.50 for output. Anthropic estimates that typical work costs about 75% less to run than on Haiku 4.5. Yahoo Finance notes that the price matches OpenAI’s GPT-6 Luna.
The model is in the Claude apps, in Claude Code and in Anthropic’s API. It is also on Amazon Web Services, Google Cloud and Microsoft’s cloud.
Input price per million tokens for requests up to 100,000 tokens, from Anthropic’s pricing pages.
Effort controls, and claims against an outside test
Haiku 5.5 is the first Haiku with effort controls, which set how long the model thinks. There are five levels, from low to max. Anthropic calls Haiku its fastest model at standard speeds.
Anthropic’s own tests show big gains over Haiku 4.5. On OSWorld, a test of using a computer, it says Haiku 5.5 scored 72.4% against 15.7%.
Artificial Analysis ran its own Intelligence Index. Haiku 5.5 at its highest effort scored 43, second of 182 models in its class, where the middle score is 13. It also used 440 million output tokens on the test, against a typical 100 million.
VentureBeat stresses that most results are reported by Anthropic itself. Token prices alone cannot show what it costs to finish a job, it adds.
Other price changes
Anthropic also halved the price of cache reads for Claude Sonnet 5.5, from $0.20 to $0.10 per million tokens. Cache reads are repeated parts of a prompt that the system has stored.
Subscribers to some plans now get monthly API credits: $100 for Max 5x, $200 for Max 20x and up to $500 shared by a Team. The credits work in the Claude Console, not in Claude Code.
Sources
Every fact in this story comes from the sources below. Open them to check our work.
- 1
- 2
- 3Research · October 7, 2026Anthropic launches Claude Haiku 5.5 with 90% API price reduction, matching GPT-6 Luna VentureBeat
- 4Research · October 7, 2026Anthropic reveals Haiku 5.5 model as AI pricing war intensifies Yahoo Finance
- 5
- 6Research · October 7, 2026Claude Haiku 5.5: Intelligence, Performance & Price Analysis Artificial Analysis
The prices come from Anthropic’s own pages; the 75% saving and the benchmark gains are Anthropic’s claims. The Intelligence Index score is Artificial Analysis’s own test. Silicon AI News is written with Anthropic’s Claude models.
Haiku 5.5 cuts small-model input prices by 90% and ranks second of 182 in an outside test, though it uses many tokens. How we rate the mood