Claude Opus 4.7 debuts

Anthropic launched Claude Opus 4.7 on April 16, making it the company’s most capable generally available model. (anthropic.com) Large language models predict the next token in a sequence, then chain those predictions into answers, code, or image analysis. Anthropic said Opus 4.7 improves the parts developers care about most: advanced software engineering, long-running tasks, instruction-following, and visual analysis. (anthropic.com) Anthropic said the model can work through complex coding jobs with more “rigor and consistency,” verify its own outputs before replying, and process higher-resolution images than earlier Claude models. The company also said Opus 4.7 is available across Claude apps, the Claude API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. (anthropic.com) For developers, the practical change is that Claude can now inspect denser screenshots, diagrams, and documents without shrinking them first. Anthropic’s API docs say Opus 4.7 raises maximum image resolution to 2,576 pixels or 3.75 megapixels, up from 1,568 pixels or 1.15 megapixels in earlier Claude models. (claude.com) Anthropic kept pricing unchanged from Opus 4.6 at $5 per million input tokens and $25 per million output tokens. The API model ID is `claude-opus-4-7`, and the model supports a 1 million token context window and 128,000 maximum output tokens. (anthropic.com, claude.com) The release lands nine days after Anthropic introduced Claude Mythos Preview on April 7, a separate model the company said was especially strong at cybersecurity work. Anthropic said Mythos Preview would stay limited while it tested new safeguards on a less capable public model first. (anthropic.com, anthropic.com) That made Opus 4.7 the test case. Anthropic said it deliberately reduced some cyber capabilities during training and added systems that “automatically detect and block” prohibited or high-risk cybersecurity requests, while inviting vetted researchers into a Cyber Verification Program for legitimate security work. (anthropic.com) Anthropic also shipped product changes around the model. Its release notes say the company launched Claude Design on April 17, a new Anthropic Labs tool for making visual outputs such as prototypes, slides, and one-pagers with Claude. (claude.com) The company’s own framing is narrower than a general “smartest model” race. Anthropic said Opus 4.7 is less broadly capable than Mythos Preview, but stronger than Opus 4.6 on coding, vision, multi-step work, and professional tasks that need steadier execution. (anthropic.com) So the launch is less about a new chatbot button than a new default for paid Claude users and API customers. Anthropic is selling Opus 4.7 as a model that can take on harder coding and document-heavy work without changing the price. (anthropic.com, aws.amazon.com)