Anthropic Rolls Out Claude Opus 4.7, A Step Below Mythos

Anthropic announced the release of its new AI model, Claude Opus 4.7, at the AI Impact Summit in New Delhi. The company clarified that while Opus 4.7 represents an improvement over previous versions, it is significantly less capable than its most advanced offering, Claude Mythos Preview.

Claude Opus 4.7 Launch Details

At the AI Impact Summit on February 19, 2026, Anthropic CEO Dario Amodei introduced Claude Opus 4.7. The model is positioned as an upgrade from the earlier Claude Opus 4.6, but the company explicitly noted its performance gap compared to its flagship product.

The Flagship: Claude Mythos Preview

Anthropic's most powerful model remains Claude Mythos Preview. This model is highly specialized and excels in advanced cybersecurity applications, particularly in identifying vulnerabilities and security flaws within software.

• Deployment: Mythos Preview was announced earlier this month and is currently rolling out to a select group of companies.
• Initiative: Its deployment is part of a specialized cybersecurity program known as Project Glasswing.

Safety and Guardrails

To manage the deployment of Opus 4.7 and guide future development, Anthropic integrated strict safety measures into the new release. The company stated that the model includes automatic safeguards designed to prevent misuse.

These safeguards are intended to:

• Automatically detect and block requests that indicate prohibited use.
• Block requests related to high-risk cybersecurity activities.

Anthropic emphasized that the data gathered from the real-world deployment of these safeguards will be crucial for the eventual goal of broadly releasing Mythos-class models to a wider audience.