Anthropic: Claude Haiku 5.5 (batch)

Claude Haiku 5.5 (batch) is a text model from Anthropic, offered through api14 under the model id anthropic/claude-haiku-5.5:batch. It is the successor to Claude Haiku 4.5, and Anthropic positions it as a small, fast model aimed at high-volume, cost-sensitive work. The batch variant is the version you call through api14's OpenAI-compatible endpoint.

The model accepts text, images, and files as input, and it returns text. Its context window is 1,000,000 tokens, which is large enough to hold long documents, extended conversation histories, or sizable codebases in a single request. The maximum output is 128,000 tokens, so responses can be long when the task requires it. Anthropic describes the model as stronger than its predecessor in coding, computer use, and related agentic tasks.

Those capabilities point to a fairly specific set of uses. Summarization is a natural fit, especially over long inputs that would otherwise need chunking. Subagents are another: a fast model that can be called many times in parallel suits orchestration patterns where a larger model plans and Haiku-class models execute. Browser use and computer use workloads also follow from the model's stated strengths, as does coding assistance where latency and throughput matter more than maximum reasoning depth. Because images and files are accepted as input, document and screenshot analysis can be handled in the same request as the text prompt.

To call it through api14, use the model id anthropic/claude-haiku-5.5:batch with the OpenAI-compatible chat completions interface. The supported request parameters are include_reasoning, max_tokens, reasoning, reasoning_effort, response_format, stop, structured_outputs, tool_choice, tools, and verbosity. Tool calling and structured outputs are both available, which means the model can be wired into function-calling loops or constrained to emit JSON-shaped responses. Reasoning can be requested and its effort level tuned, and the include_reasoning flag controls whether reasoning content is returned.

The batch designation is worth noting when you are choosing between this and other models on the platform. If your workload is interactive and needs the lowest possible latency, or if it demands the deepest reasoning available, a different model may suit you better. For high-volume, cost-sensitive jobs where the input is long and the output is bounded, this model is the intended fit.