Anthropic released Claude Haiku 5.5 on October 7 for high-volume tasks such as classification, summaries and agent subtasks. Its Claude API rate rises fivefold when a request exceeds 100,000 input tokens.

The model supports a one-million-token context window. But that capacity does not come at the same price throughout: developers handling long documents or agent histories need to account for the higher tier.

Where the price changes

Anthropic lists these Claude API rates:

  • Up to 100,000 input tokens: $0.10 per million input tokens and $0.50 per million output tokens.
  • Above 100,000 input tokens: $0.50 per million input tokens and $2.50 per million output tokens.

Anthropic’s pricing documentation says the prompt total includes tokens from cache reads and writes. Each request is priced separately.

VentureBeat described the launch as a 90% API price reduction. That comparison fits the lower tier’s per-token rates against Haiku 4.5, which cost $1 per million input tokens and $5 per million output tokens. The higher Haiku 5.5 tier is half Haiku 4.5’s listed rate. These are list-price comparisons, not estimates of the total cost of completing a task.

Token counts can change the savings

Anthropic says Haiku 5.5 costs about 75% less to run on average than Haiku 4.5. That is the company’s estimate, not an independent benchmark.

Anthropic’s documentation also says Haiku 5.5’s tokenizer counts the same text as approximately 30% more tokens than Haiku 4.5’s. That can affect the bill and makes a price-per-token comparison an incomplete guide to the cost of a task.

The model is available through the Claude API, Amazon Bedrock, Google Cloud and Microsoft Foundry. The rates above apply to the Claude API; partner platforms may set different prices. Developers should check both request length and platform pricing before estimating savings.

Haiku 5.5’s low Claude API rates apply below 100,000 input tokens. Longer requests enter a tier priced five times higher per token, while Anthropic’s tokenizer can count the same text differently from its predecessor.