Why did Claude Haiku 5.5 just get 90% cheaper — and what can I actually do with it?

Haiku 5.5 targets high-volume summarizing, classification, document Q&A, and subagent work at much lower cost, with important token and long-prompt caveats.

Anthropic released Claude Haiku 5.5 on Oct. 7 at about 90% below Haiku 4.5 pricing: $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens, with a one-million-token context window. The reason is a shift in where small models create value: they are now capable enough for high-volume work such as summarizing, classifying, document Q&A, and delegating narrow subagent tasks, so competition is moving from maximum benchmark scores to cost per completed job.

For repetitive text chores, pick Haiku rather than a flagship model, group non-urgent requests into batch jobs for another 50% reduction, and use prompt caching when the same instructions or reference document repeat; cached input can fall to $0.01 per million tokens. Budget carefully: Anthropic’s new tokenizer may count about 30% more tokens, and rates rise fivefold once a prompt crosses 100,000 tokens. Split long document collections into focused chunks, retrieve only the passages needed, and measure the full task cost—not just the advertised base rate.