Anthropic: Claude Haiku 4.5

Claude Haiku 4.5 is a text generation model from Anthropic, released by the provider on 15 October 2025. It sits in Anthropic's Claude family as the fastest and most efficient option, positioned to deliver near-frontier intelligence at lower latency than the larger Claude models. If you already route Anthropic models through your stack, this is the small, quick member of that lineup rather than a separate lineage.

On input, the model accepts text, images and files, and it produces text only. The context window is 200,000 tokens, which is enough to hold long documents, extended conversations or a mix of prose and screenshots in a single request. The maximum output is 64,000 tokens, so long-form generations and detailed structured responses are both within range. Because images and files are accepted as input, you can send a PDF or a screenshot alongside your instructions and ask for a written answer, without a separate extraction step in front of the model.

Those capabilities point to a fairly specific set of jobs. Document review and summarisation over long files, question answering grounded in a pasted knowledge base, and extraction of structured data from mixed text and image inputs are all natural fits. The same profile suits chat assistants and agent loops where you want a capable model that returns quickly, and classification or routing steps that sit in front of a heavier model. Tool use and structured outputs make it usable as the reasoning component in a pipeline rather than just a chat endpoint.

Through api14 you call it with the model id "anthropic/claude-haiku-4.5" on the OpenAI-compatible API, so existing OpenAI SDK clients work with a changed base URL and model string. The request parameters it supports are include_reasoning, max_completion_tokens, max_tokens, reasoning, response_format, stop, structured_outputs, temperature, tool_choice, tools, top_k and top_p. In practice that covers sampling control (temperature, top_p, top_k), output length (max_tokens, max_completion_tokens), stopping conditions, JSON and schema-constrained responses via response_format and structured_outputs, and function calling via tools and tool_choice. The reasoning and include_reasoning parameters let you opt into reasoning behaviour and decide whether that reasoning is surfaced.

The main thing to weigh is scope. This is a text-output model, so anything requiring generated images, audio or video needs a different model. It is also the efficiency tier of the Claude family, so if a task genuinely needs the strongest reasoning Anthropic offers, a larger Claude model is the more appropriate choice and Haiku 4.5 is better used where speed and volume matter more than peak capability.