Anthropic released Claude Opus 4.7, an updated version of its artificial intelligence model that offers improvements in software engineering and image processing capabilities, according to a company statement.

The model features enhanced vision processing, accepting images up to 2,576 pixels on the long edge, representing approximately 3.75 megapixels. This represents more than three times the resolution capacity of previous Claude models.

Opus 4.7 demonstrates improved instruction-following capabilities compared to its predecessor, Opus 4.6. The company notes this enhanced precision may require users to adjust existing prompts, as the new model interprets instructions more literally than previous versions.

The model includes cybersecurity safeguards that automatically detect and block requests indicating prohibited or high-risk cybersecurity uses. These safeguards serve as testing ground for eventual broader release of the company's more capable Mythos-class models. Security professionals can access cybersecurity features through Anthropic's new Cyber Verification Program.

Opus 4.7 is available through Claude products, the company's API, Amazon Bedrock, Google Cloud's Vertex AI, and Microsoft Foundry. Pricing remains unchanged from Opus 4.6 at $5 per million input tokens and $25 per million output tokens.

The release introduces an "xhigh" effort level option, positioned between existing high and maximum effort settings. This provides users additional control over the balance between reasoning capability and response speed.

The model uses an updated tokenizer that may increase token usage by approximately 1.0 to 1.35 times depending on content type. The company has published a migration guide for users transitioning from Opus 4.6.

Anthropic also launched additional features including task budgets in public beta for API users and an ultrareview command for Claude Code that identifies bugs and design issues.