Claude Haiku 5.5

Claude Haiku 5.5 shipped on October 7, 2026 as claude-haiku-5-5, the fourth model in the Claude 5.5 family and the one Anthropic flagged as coming when Sonnet 5.5 launched. The price card is the thing to read carefully, because it is the first Claude model whose rate changes with prompt length rather than with the model you pick. Prompts up to 100,000 tokens bill at $0.10 per million input tokens and $0.50 output, with cache reads at $0.01, a five-minute cache write at $0.125 and a one-hour write at $0.20. Cross 100,000 tokens and every rate on that request multiplies by five, to $0.50 and $2.50 with cache reads at $0.05. Batch is half off either way. Against Claude Haiku 4.5 at $1 and $5 the short-prompt rate is a 90 percent cut on both sides, and Anthropic puts the average saving nearer 75 percent per task because 5.5 uses the newer tokenizer introduced with Opus 4.7, which produces roughly 30 percent more tokens for the same text. The capability jump is a bigger step than the Haiku line usually takes: the context window goes from 200K to 1 million tokens, max output from 64K to 128K, and it is the first Haiku with adaptive thinking and an effort parameter, defaulting to medium. Input is text and images, the knowledge cutoff is June 2026, and the minimum cacheable prompt is 512 tokens. Two migration notes: temperature, top_p, and top_k return a 400 error, and thinking blocks only work in the account that produced them or an account linked to it. Anthropic positions it for classification, extraction, routing, and subagent work rather than open-ended reasoning, and Haiku 4.5 is still active on the price list at $1 and $5. It is available on the Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry, and Claude Platform on AWS, with retirement no sooner than October 7, 2027.