Anthropic: Claude Opus 4.7

Claude Opus 4.7 is a text model from Anthropic, the next generation of the company's Opus family. It was released by the provider on 16 April 2026 and is built for long-running, asynchronous agents. It builds on the coding and agentic strengths of Opus 4.6, so if you have used earlier Opus models for developer tooling or agent workflows, this is the direct successor in that line.

The model accepts text, images and files as input, and it returns text. Its context window is 1,000,000 tokens, with a maximum output of 128,000 tokens. That combination of a very large input window and a large output budget means you can hand it long documents, codebases, transcripts or mixed text and image material in a single request, and still get back a substantial response rather than a short summary. Output is text only, so anything you need in another format has to be produced as text and converted downstream.

Those capabilities point to a fairly specific set of use cases. Long-running asynchronous agents are the stated design target, which fits workflows where the model works through a multi-step task over time, calls tools, and reasons about intermediate results. The large context window suits analysis over big inputs: reviewing a large repository, working through lengthy contracts or reports, or answering questions over a document set you supply as files or images. Because it supports tools and structured outputs, it also fits pipelines where the model must call functions and return machine-readable results rather than free-form prose.

Through api14 you call it with the model id "anthropic/claude-opus-4.7" on the OpenAI-compatible API, so existing OpenAI-style clients work with a change of base URL and model id. 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 means you can control output length, steer reasoning depth and effort, constrain responses with response formats or structured outputs, define and select tools, set stop sequences, and adjust verbosity.

One thing to weigh is that this is a text-output model, so tasks that need image, audio or other non-text output are not a fit. If your workload is short, simple and latency-sensitive, the very large context window and reasoning controls may be more than you need, and a smaller model in the catalogue could serve you better. Where the model earns its place is long-context, tool-using, asynchronous agent work and heavy analysis over large inputs.