Anthropic released Claude Haiku 5.5 on October 7, 2026, positioning the small model for high-volume, cost-sensitive tasks such as summarization, classification and repetitive coding support. The company says it is available on its platform and through AWS, Google Cloud and Microsoft Azure.
The API model identifier is claude-haiku-5-5. Anthropic's published pricing distinguishes prompts up to 100,000 tokens from larger prompts. In the lower tier, input costs $0.10 per million tokens and output costs $0.50 per million. Above 100,000 prompt tokens, those rates are $0.50 and $2.50 respectively.
Cache-read prices are $0.01 per million tokens in the lower tier and $0.05 in the upper tier; cache writes are $0.125 and $0.625. These are token prices, not a flat subscription price or a guaranteed cost for every task. Prompt size, output length and caching determine a request's bill.
Haiku 5.5 adds an adjustable effort setting. Anthropic says users can trade off cost and capability, and describes stronger benchmark performance than Haiku 4.5. Those are vendor-run evaluations; we have not reproduced them, and they do not guarantee performance on a particular workload.
The announcement also cuts Sonnet 5.5 cache reads from $0.20 to $0.10 per million tokens. Anthropic estimates this lowers costs by around 20 percent on most agentic work, but that estimate depends on how much of a workload uses cached input.
Anthropic says larger Sonnet and Opus models remain better choices for complex coding tasks in its evaluations. Developers comparing Haiku should therefore test quality and full request cost, rather than read the lowest token rate as proof it is best for every task.