Anthropic Releases Claude Opus 5 with Enhanced Cybersecurity Features

On July 25, 2026, Anthropic launched its AI model Claude Opus 5, which claims to match the cybersecurity vulnerability detection capabilities of its predecessor, Mythos 5. While Opus 5 can identify vulnerabilities effectively, it significantly lags in exploiting them. The model is designed for everyday use, especially in knowledge work and coding, and is priced at $5 per million input tokens and $25 per million output tokens, making it half the cost of Mythos 5. The release follows heightened scrutiny from the US government regarding cybersecurity safeguards, leading to stronger protections compared to Opus 4.8. Anthropic emphasizes that Opus 5 is the most aligned model yet, with improved safeguards against misuse. The launch comes amid competitive pressures from other AI firms, particularly those offering cheaper alternatives. Despite its capabilities, Opus 5 is not positioned as a state-of-the-art tool for risky cybersecurity tasks.

Key Points: • Claude Opus 5 launched on July 25, 2026, with improved cybersecurity features. • The model matches Mythos 5 in vulnerability detection but lags in exploiting vulnerabilities. • Opus 5 is priced at $5 per million input tokens, making it half the cost of Mythos 5.