Anthropic positioned Claude Opus 5 as an improvement in completed work for the same 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 → prices as Opus 4.8. Its July 24, 2026 announcement reports more than twice its predecessor's performance on Frontier-Bench v0.1, a software-engineering 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 →, at lower cost per task.
The announcement attributes this to the model working more efficiently, and says customers can use its effort setting to favor intelligence or conserve tokens for faster, cheaper results. At launch, input and output prices were $5 and $25 per million tokens. Because the base price is unchanged, any lower cost per task is the company's reported result on the evaluated tasks, not a general guarantee for every workload.
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.
