Anthropic released Claude Haiku 5.5 on October 7, describing it as "the cheapest, fastest, and most capable small model we've ever released." The company says the model targets high-volume, cost-sensitive work — summaries, compaction, database queries and classification — and pairs with the larger Opus 5.5 and Sonnet 5.5 as a subagent on coding tasks. On price, Anthropic states Haiku 5.5 "now costs around 75% less to run" than Haiku 4.5, while halving Sonnet 5.5's cache-read price makes that model about 20% cheaper on most agentic work. It is the first Haiku-class model with an adjustable effort setting, and Anthropic is adding a monthly API credit for Claude Max and Team subscribers. Published benchmarks include GDPval-AA v2.1 at 1620 (versus 735 for Haiku 4.5), OSWorld 2.1 at 72.4% (versus 15.7%), and Humanity's Last Exam at 45.9% without tools.
Our takeOur take: This is a blunt cost-efficiency play, and the benchmarks Anthropic chose — agentic and computer-use tasks — are exactly where cheap models used to falter. If the ~75% price cut holds in production, Haiku-class models become a viable default workhorse rather than a fallback. Two caveats: the numbers are vendor-published and measured largely against Anthropic's own predecessor, and the bundled subscriber credit suggests the real goal is pulling users into paid tiers.