Anthropic has launched Claude Haiku 5.5, a new version of its smallest and cheapest model, aimed at high-volume, cost-sensitive tasks such as summarization, classification, database queries and live customer support. The company says it is the fastest and most capable small model it has released, and that on average it costs around 75% less to run than Claude Haiku 4.5, the model it replaces about a year later.
The release arrives one month after Anthropic upgraded its larger Opus and Sonnet models, and it lands in the middle of an AI price war. Anthropic also halved the cost of cache reads for Claude Sonnet 5.5, which it says makes that model around 20% cheaper for most agentic work, and introduced monthly API credits for Claude Max and Team subscribers.
Claude Haiku 5.5 specs and benchmarks
The benchmark numbers show a wide gap between the new model and its predecessor. According to Anthropic's published results, Claude Haiku 5.5 reaches 72.4% on the OSWorld-2.1 computer-use test, up from 15.7% for Haiku 4.5, and scores 1,620 on the GDPval-AA v2.1 benchmark for real-world professional work, more than double the 735 of the older model.
Haiku 5.5 is also the first Haiku-class model with adjustable effort controls, letting users choose between low, medium, high, xhigh and max settings to balance cost against intelligence. Reasoning cannot be disabled, and the model defaults to medium. Anthropic says it pairs well with Opus 5.5 and Sonnet 5.5 as a subagent, but that the larger models remain the better choice for complex coding work.
- OSWorld-2.1 (computer use): 72.4%, up from 15.7% for Haiku 4.5
- Humanity's Last Exam: 45.9% without tools and 57.4% with tools, up from 10.2% and 18.7%
