Anthropic launched Claude Haiku 5.5 on Wednesday, dropping per-token prices by as much as 90% and introducing effort controls to its smallest model for the first time1,2.
The company set pricing for requests under 100,000 tokens at $0.10 per million input tokens and $0.50 per million output tokens, down from $1/$5 on Claude Haiku 4.5. Larger requests carry a $0.50/$2.50 rate, compared with a flat $1/$5 on the prior model. Anthropic said roughly 90% of requests to Claude Haiku 4.5 fell into what is now the lower-priced tier, and estimated average savings of about 75% after accounting for a new tokenizer that uses slightly more tokens per task.
Haiku 5.5 is the first Haiku model to ship with effort controls, defaulting to medium, which let developers regulate how many tokens the model spends on a given task. Anthropic said the model is available on the Claude Platform, Amazon Web Services, Google Cloud, and Microsoft Azure, accessible via the identifier claude-haiku-5-5.
Anthropic described Haiku 5.5 as the "cheapest, fastest, and most capable small model" it has released, designed for high-volume, cost-sensitive workloads such as summaries and classification. The release is the first update to the Haiku line in nearly a year.
Haiku 5.5 is Anthropic's third 5.5-series model in a month. Fable 5.5, the remaining model in the series, is still pending and is expected to undergo a longer review process.
ANALYSIS The 90% price cut on the most common request tier positions Haiku 5.5 as a direct competitor to the cheapest inference-optimized models from other providers, compressing margins in the segment of the market that serves classification, routing, and summarization at scale.