Claude Haiku 5.5

Claude Haiku 5.5 is Anthropic’s small model for summaries, classification and routine coding work. Its prices depend on whether a prompt exceeds 100,000 tokens.

Claude Haiku 5.5 is an AI model for short, repeated tasks such as summarising documents, sorting requests and helping with routine coding. Anthropic released it on 7 October 2026. Developers can use it through Anthropic and supported cloud platforms.

Haiku is the smaller model in the Claude family. An adjustable effort setting lets a user trade response speed and cost against more reasoning. A larger model can assign bounded jobs to Haiku while handling the broader task itself.

The price changes above 100,000 prompt tokens

Haiku 5.5’s published pricing has two tiers. Tokens are the pieces of text a model processes; input is supplied to the model and output is generated by it.

Charge per million tokens	Prompt up to 100,000 tokens	Prompt over 100,000 tokens
Input	$0.10	$0.50
Output	$0.50	$2.50
Cache read	$0.01	$0.05
Five-minute cache write	$0.125	$0.625

The threshold is important when estimating a bill. Reusing a long document through the cache also has its own charge.

Security controls depend on the task

Anthropic says Haiku 5.5 permits a wider range of defensive security tasks than the default safeguards on Sonnet 5.5, while blocking penetration testing. Its launch describes stronger safeguards than Haiku 4.5. These are the provider’s access rules; organisations should check the terms for their intended work.

The same launch reduced Sonnet 5.5 cache reads to $0.10 per million tokens. Haiku’s low input price and Sonnet’s cache reduction are separate changes.

The price changes above 100,000 prompt tokens

Security controls depend on the task

More in Large Language Models