Anthropic's Claude Opus 5 Matches Mythos 5 in Cybersecurity Vulnerability Detection

Huo Xing Finance reports that on July 25, Anthropic released its next-generation AI model, Claude Opus 5, stating that its ability to identify cybersecurity vulnerabilities is nearly equivalent to that of the company’s more advanced Mythos 5 model. Anthropic said Claude Opus 5 is a general-purpose model with improved performance in tasks such as self-auditing, scientific research, and code optimization. The company noted that while Claude Opus 5 approaches Mythos 5 in vulnerability detection, it still “significantly lags” in actually exploiting those vulnerabilities. Additionally, Anthropic emphasized that Claude Opus 5 performs close to its state-of-the-art Fable 5 model across multiple capability categories, but at approximately half the cost, and is expected to become the default choice for enterprises handling routine office tasks. As companies continue to focus on AI investment costs, multiple AI firms—including OpenAI, Meta, and SpaceX AI—are accelerating competition toward more cost-efficient models. Anthropic stated that market emphasis on AI cost-effectiveness and practical application capabilities is steadily increasing.