Anthropic releases Claude Opus 5, with performance close to Fable 5 but at half the price

PANews July 25 news, according to Cryptobriefing, Anthropic released Claude Opus 5 on Friday, positioned as an efficient model for advanced programming, AI agents, and professional knowledge work. Opus 5 is priced at $5 per million tokens for input and $25 per million tokens for output, on par with Opus 4.8; its performance is close to Claude Fable 5 but its price is only half. The model has been launched on the Anthropic platform and the Claude API, as the new default model for Claude Max subscription users.

On the Frontier Bench v0.1, Opus 5 performs more than double versus Opus 4.8. In the CursorBench 3.2 test, the gap from the peak of Fable 5 is within 0.5%. The model also shows improvements in life sciences evaluations: its internal organic chemistry benchmark is 10.2 percentage points higher than Opus 4.8, and protein function tasks are up 7.7 percentage points. In terms of security, Opus 5 still lags behind the restricted Mythos 5 model on adversarial network security, and includes safeguards designed to prevent penetration testing and vulnerability exploitation. Anthropic also introduced a Fast mode, which is about 2.5 times faster than the standard version and priced at twice the cost.