Anthropic releases Claude Opus 5: near-Fable 5 performance at half the price

In the AI race there is the contest for the smartest model, and then there is a quieter but arguably more important contest: who brings that intelligence to a price that makes everyday use sensible. Claude Opus 5, released July 24, plays the second game.

Anthropic's proposition is easy to summarize: nearly all the capability of its flagship Fable 5 — within half a percentage point on some coding benchmarks — at half the cost per task. Like getting the higher class's engine in a car that burns half the fuel.

For Claude subscribers, Opus 5 becomes the default model on the Max plan, with extended-thinking reasoning enabled out of the box: the model reflects before answering without being asked to. For developers and enterprises, the $5 per million input tokens and $25 per million output tokens price list makes workloads viable that full frontier models kept prohibitive.

The trend this launch confirms: six-months-ago frontier intelligence rapidly becomes today's mid tier, and the real competition has shifted to capability-per-dollar rather than absolute capability.