Anthropic officially launched Claude Opus 4.7 on April 16, 2026, presenting it as its most capable generally available AI model so far. The release builds on Opus 4.6, which arrived only two months earlier, and is aimed at software engineering, long-horizon task execution, and complex financial analysis as the industry moves deeper into agentic workflows.
Sharper benchmark performance
According to Anthropic’s published metrics, Claude Opus 4.7 made notable gains in coding. The model scored 87.6% on SWE-bench Verified, up from 80.8% for version 4.6. Anthropic also said the model outperformed OpenAI’s GPT-5.4 and Google’s Gemini 3.1 Pro in selected tool-use and computer-interaction evaluations.
Vision was another major area of improvement. Claude Opus 4.7 can now process images up to 2,576 pixels on the long edge, roughly triple the previous resolution limit. That expansion should help the model interpret charts, user interfaces, and technical diagrams more effectively, though higher-resolution inputs also increase token consumption.
Built for developers and finance professionals
Inside the Claude Code environment, Anthropic added a new feature called /ultrareview, allowing pro and max-tier users to run multi-agent sessions for bug detection and software design review. The update reflects a broader push to make AI systems more useful in collaborative and semi-autonomous development workflows.
Anthropic also highlighted progress in financial reasoning. On the General Finance module, Claude Opus 4.7 posted a score of 0.813, compared with 0.767 for the previous version. That suggests stronger performance in economic modeling and more rigorous financial analysis tasks.
Costs and safeguards remain central
For developers, improved capability comes with trade-offs. Some early testers reported that the updated tokenizer can use up to 1.35x more tokens for the same input. Anthropic kept pricing unchanged at $5 per million input tokens and $25 per million output tokens, while adding a public beta task budget feature to help users control spending during long autonomous runs.
On safety, the company said the model includes new automated protections against high-risk cybersecurity misuse and has improved resistance to harmful content generation. Claude Opus 4.7 is now available through the Claude API, Amazon Bedrock, Google Vertex AI, and Microsoft Foundry, while retaining the previously introduced 1 million-token context window.