Anthropic has launched Claude Opus 5, a new AI model that matches the performance of its Claude Fable 5 model on several coding and knowledge-work evaluations at half the cost. Opus 5 more than doubled Opus 4.8’s performance at a lower cost per task. The model achieved state-of-the-art results on coding and knowledge-work evaluations, including Frontier-Bench and GDPval-AA, while remaining behind Mythos 5 on cybersecurity tasks. Unlike Fable 5 and Mythos 5, Opus 5 is not subject to a 30-day data retention policy. Anthropic said the model, like Opus 4.8, does not have data retention requirements for general access. The model improved on scientific research tasks, including structural biology, organic chemistry and bioinformatics, but remains behind Mythos 5 in biology research and offensive cybersecurity capabilities. Although it can identify vulnerabilities at a level similar to Mythos 5 in some evaluations, it is substantially less capable of turning those vulnerabilities into exploits. The company said Opus 5 has safeguards that allow some cybersecurity and biology-related research while restricting activities such as binary-based vulnerability scanning, penetration testing and exploit generation. Developers can access the model through the Claude API using the model name claude-opus-5. The model is now the default option for Claude Max users and the strongest model available on Claude Pro.