Claude Opus 5 của Anthropic đạt mức độ phát hiện lỗ hổng bảo mật ngang bằng Mythos 5

The Huoxing Finance report states that on July 25, Anthropic released its next-generation AI model, Claude Opus 5, claiming that its ability to detect cybersecurity vulnerabilities is nearly comparable to the company’s more advanced Mythos 5 model. Anthropic stated that 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 although Claude Opus 5 approaches Mythos 5 in vulnerability detection capability, it still “significantly lags” in actually exploiting these vulnerabilities. Additionally, Anthropic emphasized that Claude Opus 5 performs close to its most advanced 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 companies—including OpenAI, Meta, and SpaceX AI—are accelerating competition to develop more cost-efficient models. Anthropic stated that market emphasis on AI cost-effectiveness and practical application capabilities is steadily increasing.