Claude Opus 5's launch emphasized the cost of completing work, with an effort setting that lets users balance more reasoning against response time and cost. In its July 24, 2026 announcement, Anthropic reported improved performance over Opus 4.8 at the same per-tokenThe chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.Full definition → price and presented results across effort levels.
The useful implication is that the model name alone does not specify the operating tradeoff: the effort setting also affects how much work the model performs before answering. These were launch-era evalA repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.Full definition → claims. The current model reference lists Opus 5 as still available but superseded within the Opus line by 5.5.
Terms in this piece · Glossary
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.