Anthropic: Claude Opus 4.5 (batch)

Claude Opus 4.5 (batch) is a text model from Anthropic, identified on api14 as anthropic/claude-opus-4.5:batch. Anthropic describes Claude Opus 4.5 as its frontier reasoning model, optimized for complex software engineering, agentic workflows, and long-horizon computer use. The batch variant is the same model served through a batch endpoint, which suits workloads where you can submit many requests together rather than needing an answer immediately.

The model accepts text, images, and files as input, and it returns text. Its context window is 200,000 tokens, and the maximum output is 64,000 tokens. That combination of a large context and a large output budget means you can feed in long documents, codebases, or image-heavy material and still get back a substantial response in one call. Supported parameters include temperature, max_tokens, stop, response_format, structured_outputs, tools, tool_choice, reasoning, include_reasoning, and verbosity, so you can control sampling, constrain output shape, and enable tool use or reasoning where needed.

Those capabilities point to a fairly specific set of use cases. Complex software engineering tasks fit well: reviewing or refactoring large files, working across a repository that fits in context, or generating code with tool calls to inspect and modify a project. Agentic workflows are another fit, since tool_choice and tools let the model decide when to call functions, and reasoning support helps on multi-step problems. Long-horizon computer use, as Anthropic frames it, means tasks that unfold over many steps rather than a single prompt and response. The multimodal input also helps when a task mixes screenshots, diagrams, or documents with text instructions, such as extracting structured data from a scanned form or reasoning over a chart alongside written requirements.

To call it through api14, use the OpenAI-compatible chat completions endpoint with the model id anthropic/claude-opus-4.5:batch. Because the API is OpenAI-compatible, existing OpenAI SDKs and clients work with a change of base URL and API key, and you pass the model id as the model field. The request body can include any of the supported parameters listed above: temperature, max_tokens, stop, response_format, structured_outputs, tools, tool_choice, reasoning, include_reasoning, and verbosity. If you want strict JSON or a schema-conformant object back, use response_format or structured_outputs rather than prompting for JSON by hand. If you want the model to call functions, supply tools and set tool_choice.

One practical consideration is that this is the batch variant, so it is aimed at asynchronous, high-volume jobs rather than interactive, low-latency chat. If your application needs immediate responses in a conversational loop, a non-batch serving of the same model would be the better fit. The 200,000 token context is large but finite, so extremely long histories or very large repositories may still need chunking or retrieval. Output is text only, so tasks that require generated images or audio fall outside this model.