Latest News About Opus 4.7 Xhigh

Here’s what I can share about Claude Opus 4.7 (Opus 4.7) and the new xhigh capability based on the latest reporting.

Direct answer

• Claude Opus 4.7 is Anthropic’s updated Opus model released mid-April 2026, introducing a new xhigh effort level and enhanced vision and coding abilities, alongside token-management features in beta. It also raises image input resolution to 2,576 pixels on the long edge (about 3.75 MP), and supports larger output contexts (up to 128k tokens in some reports).[3][5][8]

What’s new in Opus 4.7

• Vision and image handling: Upgraded visual input to 2,576px long edge (roughly 3.75 MP), enabling higher-fidelity analysis of diagrams, documents, and complex visuals.[5][3]
• Coding and tooling: Notable gains in coding performance, with claims of better handling of complex coding tasks and a reported ~13% improvement on a standard coding benchmark in some sources.[8][3]
• New effort level: Introduction of an xhigh effort level between high and max, intended for deeper, more thorough reasoning without stepping to the maximum computational load.[3][5][8]
• Task budgets: Beta feature allowing users to cap token usage and tool calls across multi-step tasks, helping manage costs and latency in agentic workflows.[8][3]
• Self-verification and safety: Reports note enhanced self-checking and output verification, aiming to reduce errors in long-horizon tasks.[3][8]
• Availability and integration: Opus 4.7 is being rolled out across Anthropic products and via major cloud and developer platforms, with pricing reportedly unchanged at $5 per 25,000 tokens or similar publicized tiers in several outlets.[5][8][3]

Context and comparisons

• Anthropic themselves have framed Opus 4.7 as a meaningful upgrade, while acknowledging it does not yet reach the capabilities of Mythos, their unreleased advanced system, which is kept private for safety reasons.[7]
• Market coverage notes: Reports indicate Opus 4.7 is being integrated with platforms like GitHub Copilot, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry, signaling broad adoption across developer tooling ecosystems.[8][3]

Illustrative example

• Use case: A software engineer uses Opus 4.7 for a long-horizon coding task that involves analyzing a complex data diagram, writing and testing code, and generating documentation. The xhigh setting allows deeper reasoning steps, Task Budgets help cap token use across the entire workflow, and the 3x vision resolution helps the model interpret the diagram accurately while producing documented code and explanations in one pass.[5][3][8]

If you’d like, I can pull more focused details (e.g., pricing specifics, exact token limits, or platform availability) and summarize them in a concise spec sheet. I can also compare Opus 4.7 to Opus 4.6 on the key dimensions you care about (vision, coding, latency, and cost).