Claude Sonnet 5 increases performance at a higher cost per task

Claude Sonnet 5 scored 53 on the Artificial Analysis Intelligence Index, placing it 5th in the rankings, trailing GPT-5.5 (xhigh) and Opus 4.8 (max) by only 2-3 points. In max mode, this represents a 6-point gain over Sonnet 4.6.

The model is more resource-intensive: it generates 40% more output tokens per task than Sonnet 4.6 and performs three times as many agentic steps on the AA-Briefcase and GDPval-AA benchmarks. Consequently, a single Index task costs $2.29, which is roughly double the cost of Sonnet 4.6 and 15% higher than Opus 4.8.

Token pricing remains $3 per million input tokens and $15 per million output tokens, with a discount to $2 and $10 available until September 1. The model supports a 1M token context window and introduces an xhigh mode. On agentic tasks, Sonnet 5 performs on par with or slightly better than Opus 4.8, though it lags behind larger models on complex physical benchmarks like CritPt (17%).