Claude Haiku 5.5

Claude Haiku 5.5 is a text chat model built by Anthropic. It belongs to the Haiku branch of the Claude family, which Anthropic positions as its compact, latency-sensitive tier — sitting below the Sonnet and Opus lines that are aimed at heavier reasoning and long-horizon agentic work. Haiku models are generally selected when a workload needs many requests handled quickly and consistently rather than the deepest possible reasoning on each one. On throughput, independent measurements from Artificial Analysis put Claude Haiku 5.5 at roughly 137.8 output tokens per second, with a measured time to first token of about 4,697 ms. That combination is worth reading carefully: once generation starts, tokens arrive at a brisk rate, but the measured initial latency is substantial, which matters more for short interactive turns than for long generations where the streaming rate dominates total completion time. Other specifications for this entry — context window, image input, and tool-calling details — are not tracked in our database, so we do not list them here; check Anthropic's model documentation for the authoritative specification. In practice, Haiku-tier Claude models are used for high-volume, cost-sensitive text work: classification, extraction, summarization, routing, content tagging, and as the worker model inside larger pipelines where a Sonnet- or Opus-tier model handles planning. Because Claude Haiku 5.5 is served through multiple providers, the practical decision usually comes down to the price, rate limits, and region offered by each one — compare those in the pricing table on this page.