Anthropic released Claude Haiku 5.5 on October 7, 2026. The model ID is claude-haiku-5-5. Anthropic describes it as a model for high-volume, latency-sensitive tasks such as classification, extraction, and routing.
Context window: 1M tokens. Max output: 128K tokens.
Prompts up to 100,000 tokens: $0.10 input and $0.50 output per million tokens. Cache reads cost $0.01.
Prompts over 100,000 tokens: $0.50 input and $2.50 output per million tokens. Cache reads cost $0.05.
Adaptive thinking is on by default. The default effort is medium.
The same text counts as approximately 30% more tokens than on Claude Haiku 4.5.
Retirement: not sooner than October 7, 2027.
Prerequisites
An Anthropic API key in ANTHROPIC_API_KEY
An existing app that calls the Claude API
Steps
Change the model name in your request to claude-haiku-5-5.
Remove temperature, top_p, and top_k. The release notes say these can return a 400 error. So can budget_tokens and assistant message prefill.
Expect replies that can begin with thinking blocks. Read the text blocks, not only the first block.
Keep prompts under 100,000 tokens if you want the lower price tier.
Test it
Send one short request with the new model name. It worked if the response comes back without a 400 error and the model field shows claude-haiku-5-5.
Not confirmed
No code sample here. The pages used for this entry show none for this launch.
The official pages give no speed figures, only the label Fastest relative to the current lineup.
Amazon Bedrock and Google Cloud set their own retirement dates.
No retirement date for Claude Haiku 4.5 is stated in this announcement.
Wirestack