Anthropic Releases Claude Opus 4.7 with Autonomous Agent Capabilities

Anthropic PBC released Claude Opus 4.7 on April 16, 2026, as the high-performance tier of its AI lineup. The model introduces Auto Mode, which enables the AI to execute long-running tasks independently via browsers or terminals, shifting its role from a reactive chatbot to a collaborative digital architect. It also features high-resolution vision supporting images up to 2,576 pixels and a new xhigh effort level for controlling reasoning depth.

To balance innovation with safety, Anthropic intentionally designed Opus 4.7 with reduced cybersecurity capabilities compared to its restricted Mythos system. As part of the Project Glasswing security initiative, the model includes safeguards to automatically detect and block high-risk cyber requests, though security professionals can bypass these via a Cyber Verification Program. The company acknowledged the model is modestly weaker in responding to harmful prompts than previous versions, using it as a testing ground for safeguards intended for the future Mythos release.

The model was developed using reinforcement learning from human feedback and the company's Constitutional AI framework to improve creative writing, software engineering, and financial analysis. Opus 4.7 is available through Claude.ai, the Anthropic API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. Priority access is granted to Claude Pro subscribers for a $20 monthly fee.