Claude Haiku 5.5 Prices Short Requests at $0.10 per Million Input Tokens

Anthropic released Claude Haiku 5.5 this week at $0.10 per million input tokens and $0.50 per million output tokens for requests up to 100k tokens. Anthropic says that is roughly 90% cheaper than Haiku 4.5 on shorter requests and about 75% cheaper on average. Above 100k tokens the price is $0.50 input and $2.50 output.

The target is high-volume, cost-sensitive work: summaries, compaction, database queries, and classification. It is the first Haiku-class model with an adjustable effort setting. One catch: it uses an updated tokenizer that consumes slightly more tokens per task, so cost per completed task will not fall by exactly the headline percentage. Measure your own workload before you rewrite a budget.

Customer results give a sense of where it fits. HubSpot scored 92.8% on its CRM evaluation suite, Asana saw over 30% lower latency, and Box saw an 11-point gain at about half the latency. On benchmarks, Haiku 5.5 reports 72.4% on the offline subset of OSWorld 2.1 and 39.2% on Terminal-Bench 4.0. Anthropic is explicit that Sonnet 5.5 and Opus 5.5 remain better for complex agentic coding.

The same announcement halved Sonnet 5.5 cache-read pricing to $0.10 per million tokens. Sonnet 5.5 launched September 28 at $2 per million input tokens and $10 per million output, so the company now sells a clear ladder: a cheap tier for volume, a mid tier for production automation, and a top tier for the hardest work.

The practical move is routing. Send classification, summarization, and subagent chores to the small model, keep the harder reasoning and coding on the larger ones, and judge the split on cost per outcome shipped, not cost per token.