Anthropic has officially introduced Claude Opus 4.7, its most capable artificial intelligence model currently available to the general public. Released on April 16, 2026, this major update delivers significant enhancements in software development, high-resolution vision, and complex multi-step reasoning.
While Anthropic continues testing its highly restricted Claude Mythos model, Claude Opus 4.7 serves as the company’s flagship public offering. The new model brings a noticeable shift in performance and personality, designed specifically to handle difficult professional knowledge work with greater consistency.
Enhanced Coding and Vision Capabilities
Software engineering is a primary focus for the Claude Opus 4.7 update. Early data indicates a 13 percent improvement on coding benchmarks, alongside a notable increase in the number of resolved production tasks compared to previous iterations.
The model also introduces high-resolution image support, marking a major milestone for the Claude family. The maximum image resolution has expanded dramatically from 1.15 megapixels to 3.75 megapixels, supporting images up to 2576 pixels. This upgrade is specifically aimed at improving vision-heavy workloads, such as understanding documents, analyzing screenshots, and executing computer-use tasks.
Additionally, Anthropic has simplified how the AI interacts with images. The model now maps coordinates directly to actual pixels on a one-to-one basis. This eliminates the need for users to perform complex scale-factor math when working with visual data. However, because higher-resolution images consume more tokens, Anthropic advises users to downsample images if maximum visual fidelity is not required.
Pricing, Tokens, and Claude Code Integration
Claude Opus 4.7 is deeply integrated into Claude Code, bringing several new developer-focused tools. Users now have access to an “xhigh” effort tier, which provides a middle ground for long coding tasks without requiring maximum latency. Another notable addition is the ultrareview command, which enables parallel, multi-agent code reviews.
Developers will also benefit from a new full-screen terminal rendering interface and a one-hour prompt cache time-to-live setting. This caching update aims to reduce input costs for coding sessions that last an hour or more.
Despite these performance upgrades, the base pricing remains identical to the previous version, Claude Opus 4.6. Users will continue to pay $5 per million input tokens and $25 per million output tokens. However, the latest update includes a new tokenizer, which has led to a modest increase in overall token usage.
A New Personality and Behavioral Changes
Beyond technical specifications, Claude Opus 4.7 features a distinct shift in communication style. The model now uses a more direct and opinionated tone. It actively moves away from the warmer, emoji-heavy, and validation-focused language found in Claude Opus 4.6.
The AI also changes how it handles long, complex tasks. It now provides more regular progress updates to users throughout extended operations. Furthermore, the system is designed to spawn fewer subagents by default, though users can still steer this specific behavior through careful prompting.
The Launch of Claude Design
Shortly after releasing the new model, Anthropic Labs launched Claude Design on April 17, 2026. Powered entirely by Claude Opus 4.7, this experimental product allows users to generate polished visual work, such as slides, prototypes, and one-pagers.
Claude Design is built to help product managers, founders, and other professionals create visuals without needing a formal design background. Users simply describe their desired outcome, and the AI generates an initial version that automatically applies designated brand guidelines, including typography and colors.
Users can refine these visuals through direct requests or manual edits. The platform allows static mockups to become interactive prototypes and supports exporting projects to formats like Canva, PDF, HTML, and PowerPoint. Currently, Claude Design is available in a research preview for Claude Pro, Max, Team, and Enterprise subscribers.
Claude Opus 4.7 vs. Claude Mythos Preview
While Claude Opus 4.7 is Anthropic’s most advanced public model, the company acknowledges it is less broadly capable than the Claude Mythos Preview. Mythos is an incredibly powerful model that excels at identifying software weaknesses and security flaws.
Because Anthropic views the advanced cybersecurity capabilities of Mythos as potentially dangerous, it remains heavily restricted. Mythos is only available to a select group of early-access organizations through a cybersecurity initiative known as Project Glasswing. Before releasing Claude Opus 4.7 to the broader public, Anthropic conducted extensive testing to ensure all necessary safety guardrails were functioning correctly.
