Anthropic released Claude Haiku 5.5 on October 7, pricing the small model below Haiku 4.5 and adding an effort setting that the company says lets developers favor cost or model intelligence.
For prompts up to 100,000 tokens, Anthropic lists prices of $0.10 per million input tokens (text an app sends to the model) and $0.50 per million output tokens (text the model returns), down from $1 and $5 for Haiku 4.5. The company says that prompt range covered about 90% of requests to the previous Haiku model.
The effort setting is new to the Haiku line, but Anthropic’s launch page does not describe what it changes behind the scenes. The company pitches the model for narrow, repeated jobs and coding subagents, while saying its larger Sonnet 5.5 and Opus 5.5 models remain better suited to complex agentic coding.
Anthropic estimates average task cost is about 75% lower than Haiku 4.5. That is less than the 90% per-token price cut: the company says Haiku 5.5’s updated tokenizer uses slightly more tokens to complete a given task. The model is available across Anthropic’s platforms and cloud partners; computer-use and browser-use support in its Python and TypeScript SDKs is in beta.
Sources
Anthropic’s Claude Haiku 5.5 launch announcement
Image: Darakshanehaluddin, “Laptop image,” Wikimedia Commons, CC0 1.0 (file page). General illustration; not a Claude product image.
