Anthropic said on the 17th that it released its latest artificial intelligence (AI) model, Claude Opus 4.7.
The new model is an upgraded version over Opus 4.6, with notably stronger software engineering capabilities. Its performance has been improved to achieve high-quality results in complex coding tasks without separate, detailed guidance.
It also improves vision features with support for high-resolution images, and delivers higher completeness and creativity in professional work areas such as interface design, slides, and document writing. Instruction-following has likewise been strengthened, increasing usability across various work environments.
On the security side, a new feature automatically detects and blocks high-risk cyber-related requests. This is the first case of applying the Project Glasswing approach Anthropic proposed to an actual service.
However, Opus 4.7 is not a cybersecurity-specialized model and has more limited performance in that area than Claude Mythos Preview, which was released separately. Mythos Preview is provided on a limited basis only to cyber defense experts and key infrastructure partners.
Pricing remains the same as before: $5 per 1 million input tokens and $25 per 1 million output tokens.