Anthropic Releases Claude Haiku 5.5, Its Fastest and Cheapest Small Model

Anthropic has released Claude Haiku 5.5, completing the Claude 5.5 line as the fastest and cheapest tier alongside Opus 5.5 and Sonnet 5.5. The model targets high-volume, cost-sensitive work such as summaries, compaction, database queries, and classification, and Anthropic says it now costs around 75% less to run than Haiku 4.5. It is the first Haiku-class model with an adjustable effort setting, and it scores 1620 on GDPval-AA v2.1 against 735 for Haiku 4.5, 72.4% on OSWorld 2.1 against 15.7%, and 39.2% on Terminal-Bench 4.0 against 0%, though Sonnet 5.5 remains the better choice for complex agentic coding at 70.6%.

API pricing per million tokens for prompts up to 100k is $0.10 input, $0.50 output, $0.01 cache reads, and $0.125 cache writes, with higher rates above that threshold. Haiku 5.5 is available on AWS, Google Cloud, Azure, and the Claude Platform under the model ID claude-haiku-5-5. Anthropic also halved Sonnet 5.5 cache read pricing to $0.10 per million tokens, which it says cuts the cost of most agentic Sonnet 5.5 work by around 20%, and introduced monthly API credits for Max and Team subscribers ($100 for Max 5x, $200 for Max 20x, up to $500 pooled for Team).

Introducing Claude Haiku 5.5