Anthropic has unveiled its new artificial intelligence model, "Claude Opus 5."

CNBC reported on July 24 that Opus 5 offers higher performance than the existing model, "Claude Fable 5," in coding and knowledge-work evaluations while costing about half as much.

Usage fees are $5 per 1 million input tokens and $25 per 1 million output tokens.

Anthropic drew a line, saying Opus 5 is not its top model in areas of "dangerous dual-use capabilities" such as cybersecurity. It said it also lags behind Mythos 5 in offensive cybersecurity and biological research capabilities.

Opus 5 is a move aimed at demand from cost-sensitive corporate customers. Companies are becoming cautious about adopting high-priced models with unclear returns on investment. Anthropic is also competing with low-cost models offered by Microsoft, Amazon, Google and several Chinese startups.

Diane Penn (다이앤 펜), head of product management for Anthropic's research division, said corporate customers are focusing more on value than price. She said that even a cheaper model is not practically helpful if it cannot deliver a similar level of quality.

Anthropic has undergone intensive validation after a limited release of "Claude Mythos Preview" to some users in April. In June, it released Mythos 5 and Fable 5, a version it says is safe for general use.