Anthropic released Claude Haiku 5.5 on Oct. 7, positioning it as a smaller model for high-volume, cost-sensitive work such as summaries, database queries, classification and browser tasks. The company says it runs about 75 percent less expensively than Haiku 4.5 on average and describes it as its fastest model so far. Anthropic’s benchmark table reports a 72.4 percent score on the offline subset of OSWorld 2.1, compared with 15.7 percent for Haiku 4.5; those are company-reported tests, not independent measures of everyday performance. The announcement also cuts Sonnet 5.5 cache-read prices in half and introduces monthly API credits for some Claude Max and Team subscribers. Anthropic also publishes model evaluations through its transparency hub.
Anthropic pitches Claude Haiku 5.5 for high-volume AI work
The Oct. 7 model release emphasizes speed and lower operating cost for repetitive tasks, with benchmark claims attributed to Anthropic.
Source published 2026-10-07 · Updated October 22, 2026