Anthropic has officially announced the general availability of Claude Opus 4.7, its most advanced model for software engineering and complex reasoning. Positioned as a direct upgrade to Opus 4.6, the new model delivers double-digit gains in task success, specifically targeting autonomous coding, financial analysis, and high-resolution multimodal understanding.

State-of-the-Art Autonomous Engineering

Opus 4.7 is designed for "hand-off" engineering work—tasks that previously required constant human supervision.

• Complex Debugging: The model shows a 13% improvement on coding benchmarks, clearing 70% on CursorBench (vs. 4.6's 58%) and resolving 3x more production tasks on Rakuten-SWE-Bench.

• Self-Verification: A key feature of Opus 4.7 is its ability to catch its own logical faults during the planning phase and verify its code before reporting back.

• Durable Autonomy: In early testing with Devin, Opus 4.7 worked coherently for hours on deep investigations that other models couldn't reliably sustain.

Substantially Better Vision

The model now supports images with over 3x the resolution of previous versions, accepting up to 2,576 pixels on the long edge (~3.75 megapixels). This enables:

• Computer-Use Agents: Better reading of dense screenshots and technical diagrams.

• Precision Extraction: Improved accuracy for life sciences patent workflows and chemical structure interpretation.

Cybersecurity Safeguards & Verification

Following the recent Project Glasswing announcement, Opus 4.7 is the first model to feature built-in automated safeguards that detect and block high-risk cybersecurity requests. Anthropic has also launched a Cyber Verification Program for legitimate security professionals (red-teamers, researchers) to access the model's full defensive potential.

New Controls for Developers

• "xhigh" Effort Level: A new reasoning tier between high and max that offers finer control over the tradeoff between deep thinking and latency.

• Task Budgets (Public Beta): A way for API developers to guide Claude's token spend over long, autonomous runs.

• Ultrareview: A new command in Claude Code that performs dedicated, senior-level code reviews to flag bugs and design flaws.

Availability and Pricing

Claude Opus 4.7 is available today on the Claude API, Amazon Bedrock, Google Cloud's Vertex AI, and Microsoft Foundry. Pricing remains unchanged at $5/million input tokens and $25/million output tokens. Note that due to an updated tokenizer and deeper reasoning, token usage per task may vary compared to Opus 4.6.

Opus 4.7 marks a significant step toward AI that acts as a "true teammate," capable of managing its own workflows, pushing back on technical decisions, and delivering production-grade code with minimal oversight.