Anthropic Launches Claude Opus 4.7 with High-Resolution Vision, Task Budgets, and Faster Agentic Performance

Anthropic launched Claude Opus 4.7 on April 16, 2026, releasing its most capable generally available model to date. The model introduces several new capabilities: high-resolution vision support up to 3.75 megapixels, allowing the model to analyze detailed images and documents; a new "xhigh" effort level that allocates additional compute for demanding reasoning tasks; and task budgets for agentic loops, which give developers precise control over how many steps an autonomous agent takes before returning control to the user. A new tokenizer also improves efficiency across languages.

Benchmark results show Opus 4.7 outperforming its predecessor across multiple dimensions. The model achieves significant gains in software engineering tasks, instruction-following, and real-world agentic use cases, including computer use and multi-step tool calling. Anthropic's internal evaluations show particular improvement on long-horizon coding tasks that require maintaining context and coherence across dozens of steps — a key capability for enterprise applications. The model is positioned below the restricted Claude Mythos Preview in terms of raw capability, but is the most powerful model available for general commercial use.

Claude Opus 4.7 is available immediately across all Anthropic Claude products including claude.ai, the Anthropic API, and cloud platforms: Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Azure AI Foundry. Pricing is set at $5 per million input tokens and $25 per million output tokens. The launch comes alongside a new AI design tool for creating websites and presentations, previewed earlier in the week by The Information, further expanding Anthropic's product surface beyond pure API access into consumer-facing creative applications.