Claude Sonnet 5.5 took #3 with 410M tokens of output

To score 56 on the Artificial Analysis Intelligence Index, the max-effort version of Anthropic’s Sonnet 5.5 generated 410M output tokens. The median reasoning model in its price tier used 88M and scored 26.

That score ranks it #3 of 216 models in its class. The index bundles 10 evaluations, among them Terminal-Bench 4.0, SciCode and Humanity’s Last Exam, so it measures more than chat quality.

Speed wasn’t measured: the output-tokens-per-second field reads N/A. We’d keep an eye on the token appetite, since it’s the part of the bill you can’t negotiate.

Its list prices of $2 per 1M input tokens and $10 per 1M output tokens sit exactly at the class medians.