Claude Haiku 4.5 (latest)

Claude Haiku 4.5 (latest) continues Anthropic's Haiku line as a compact, multimodal model positioned for responsive inference rather than the deepest reasoning in the family. It accepts text, images, and PDF documents as input while producing text-only output, making it well suited to workflows where lightweight document understanding and image-aware comprehension are needed without invoking a larger flagship model. The Haiku tier traditionally emphasizes lower latency and lower cost per token, and this release maintains that profile within OpenRouter's catalog.

Beyond basic generation, the model supports reasoning, tool calling, and temperature control, along with structured output, which lets it slot into agentic pipelines and retrieval-augmented applications where it must invoke external functions or emit machine-readable responses. Attachments handling rounds out its utility for document-heavy tasks. With a very large context window, Claude Haiku 4.5 fits workloads that require ingesting long documents, multi-turn conversations with extensive history, or large code bases, while the substantial output ceiling accommodates long-form generation. In practice, it is a pragmatic choice for teams that want Anthropic-grade quality and multimodal understanding on a budget-friendly tier.