Anthropic: Claude Haiku Latest

Claude Haiku Latest is a text generation model from Anthropic, exposed on api14 under the model id ~anthropic/claude-haiku-latest. The name reflects how the endpoint behaves: rather than pinning a single frozen snapshot, it always redirects to the latest model in the Claude Haiku family. That makes it the convenient choice when you want your integration to track Anthropic's current Haiku release without changing the model id in your code each time the family is updated.

On the input side the model accepts text, images and files, and it produces text output. The context window is 1,000,000 tokens, which is large enough to hold very long documents, extended conversation histories or sizeable codebases in a single request. The maximum output is 128,000 tokens, so responses can be long as well as the prompts that produce them. Because images and files are accepted alongside text, the model can work with mixed inputs rather than plain strings alone.

Those capabilities point to a fairly broad set of practical uses. Long context plus file input suits document analysis, summarisation and question answering over contracts, reports or research material. Image input extends the same workflow to screenshots, scanned pages and diagrams that need to be read and described in text. Tool calling and structured outputs make it a reasonable fit for extraction pipelines, classification, and agent style applications where the model has to return machine readable results or invoke functions. The large output budget helps when the task is to generate substantial text rather than a short answer.

You call the model through api14's OpenAI-compatible API using the model id ~anthropic/claude-haiku-latest, which means existing OpenAI SDK clients and request shapes generally carry over with minimal change. The supported request parameters are include_reasoning, max_completion_tokens, max_tokens, reasoning, reasoning_effort, response_format, stop, structured_outputs, tool_choice, tools and verbosity. In practice that covers output length control through max_tokens or max_completion_tokens, reasoning control through the reasoning, reasoning_effort and include_reasoning options, structured and constrained output through response_format and structured_outputs, stopping conditions through stop, and function or tool use through tools and tool_choice.

Because the endpoint always points at the newest Haiku model, behaviour can shift when Anthropic updates the family, so if you need strictly reproducible output over time, a pinned model id may serve you better. The model outputs text only, so tasks that require generated images, audio or other non-text artefacts fall outside what it can return. It is also a Haiku tier model, so if a workload demands the strongest reasoning Anthropic offers, a larger model in their lineup is likely the more appropriate pick.