Claude Opus 4.7 released

Anthropic released Claude Opus 4.7 on April 16, saying the new model is now generally available for coding, agents, vision, and other complex work. (anthropic.com) Anthropic said Opus 4.7 improves on Opus 4.6 in advanced software engineering, especially on harder tasks that previously needed closer human supervision. The company also said the model is available in Claude products, its API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. (anthropic.com) A context window is the amount of text a model can keep in view at once, like the size of a worktable. Anthropic’s model documentation says Opus 4.7 supports a 1 million-token context window and up to 128,000 output tokens. (platform.claude.com) Anthropic’s docs also say Opus 4.7 adds high-resolution image support, raising the image limit to 2576 pixels and 3.75 megapixels from 1568 pixels and 1.15 megapixels. The same page says the model introduces a new “xhigh” effort setting and a beta “task budget” feature for longer agent loops. (platform.claude.com) Anthropic is pitching the release around long-running software tasks, an area where model makers have been competing on how reliably systems can plan, use tools, and check their own work over many steps. Anthropic’s launch post said Opus 4.7 “devises ways to verify its own outputs before reporting back.” (anthropic.com) For developers, the bigger change may be migration risk. Anthropic’s release notes say Opus 4.7 includes breaking changes in the Messages API, including removal of the old extended-thinking budget and rejection of non-default sampling parameters such as temperature, top_p, and top_k. (platform.claude.com) The same migration guide says reasoning text is now omitted by default unless developers explicitly opt in to a summarized display. It also says Opus 4.7 uses a new tokenizer, which can change token counts versus Opus 4.6 by as much as roughly 35%, depending on the content. (platform.claude.com) Anthropic’s model overview lists Opus 4.7 at $5 per million input tokens and $25 per million output tokens, the same pricing Anthropic said it kept from Opus 4.6. The overview also lists a 1 million-token context window and describes Opus 4.7 as the company’s most capable generally available model. (platform.claude.com) Anthropic also said Opus 4.7 is less broadly capable than its limited-release Claude Mythos Preview, but is the first model it is shipping with new safeguards aimed at blocking prohibited or high-risk cybersecurity requests. Anthropic said legitimate security researchers can apply through its Cyber Verification Program. (anthropic.com) The immediate test is whether developers treat Opus 4.7 as a drop-in upgrade or a model that needs prompt, tooling, and budget changes before production use. Anthropic’s own documentation points them to migration guidance before they switch. (platform.claude.com)