Claude leads performance benchmarks with integrated safety mechanisms

Anthropic has released an updated Claude model that outperforms competitors on nearly all evaluated benchmarks, with the advantage increasing for longer tasks. The model implements automatic safety protocols that route requests related to cybersecurity and biological research to Opus 4.8, a previous-generation version. This redirection occurs infrequently, affecting fewer than 5% of user sessions. The new release is priced at $10 and $50 on the API, depending on plan tier, and is already available through paid Claude subscriptions.