Box Evaluation Finds Claude Opus 5.5 Used About One-Third As Many Tokens As Opus 5

A Box evaluation found Claude Opus 5.5 used about one-third as many tokens as Opus 5 on complex business tasks, adding task-level evidence as Anthropic lowered model costs.

The assessment involved business tasks that required working across multiple documents. Opus 5.5 completed them 30% faster and produced deliverables 42% shorter than Opus 5, while maintaining accuracy in the evaluation.

Those results describe performance in Box’s test, not Anthropic’s total token use. A separate tracked business sample showed token volume rising faster than spending over a similar period, but it does not establish that Opus 5.5 caused the change.

Anthropic introduced Opus 5.5 on Sept. 22 and set its API price at $4 per million input tokens and $20 per million output tokens. Opus 5 cost $5 per million input tokens and $25 per million output tokens. Cache reads for Opus 5.5 cost $0.20 per million tokens, compared with $0.50 for Opus 5.

Anthropic projected that typical workloads would cost 40% less to run on Opus 5.5 and that output generation would be more than 30% faster. The model has separate prices for input, output and cached tokens, so the token rate alone does not determine a workload’s cost.

In the tracked business sample, token volume for the week ending Oct. 11 was 52.4% higher than for the week ending Sept. 13. Spending rose 23% over the same comparison. The figures cover businesses connecting provider usage and billing data to a token-spending tracker; they do not isolate Opus 5.5’s effect. The comparison was covered in an earlier look at business AI token spending.