Opus 5 is designed for everyday coding, knowledge work, and long, multi-step agentic tasks. It has become the default model in Claude Max and the most powerful model in Claude Pro. The API price is $5 per million input tokens and $25 per million output tokens, the same as Opus 4.8.
Anthropic says that Opus 5 approaches Fable 5 in quality at roughly half the cost per completed task. At the same time, it is weaker than Mythos 5 at turning discovered vulnerabilities into working exploits. The benchmark advantages mainly come from Anthropic itself and early-access partners, so they still require independent verification.
Alongside the model, two beta features were released: the ability to change the set of available tools directly during a conversation without losing the prompt cache, and automatic fallback to another model if the safety classifier blocks a request.