Claude Haiku 4.5 (latest)

Claude Haiku 4.5 sits at the efficient end of the Haiku family, positioned by Anthropic as a fast lane for lightweight agents, routine office tasks, and responsive conversational use. It accepts text and images and produces text output, while its closed-weight distribution means access is delivered through hosted APIs rather than downloadable checkpoints. The model retains the broad 200,000-token context window and 64,000-token maximum output that define current Claude deployments, giving it enough headroom for document-grounded assistants, multi-step agent loops, and longer chat sessions without sacrificing the snappy response profile expected of the Haiku tier.

In practice, the model fits workflows where latency and cost efficiency matter more than frontier reasoning depth, such as customer support triage, internal tooling, summarization passes, and orchestration of tool calls in agent pipelines. Independent listings confirm native reasoning, tool calling, temperature control, and structured output support, which makes it straightforward to drop into existing agent frameworks and JSON-constrained pipelines. With pricing well below flagship Claude tiers, Haiku 4.5 is best suited as the always-on workhorse in a tiered setup, handling high-volume traffic and lightweight reasoning while heavier or more nuanced tasks are escalated to larger Claude models.