In an independent test, Claude Sonnet 5 placed fifth with 53 points at peak performance, beating the pricier Opus 4.8 on some agent-based tasks, but its massive jump in token consumption makes the model more expensive per task than Anthropic's previous top model. On paper, Sonnet 5 keeps the same token prices as its predecessor: $3 per million input tokens and $15 per million output tokens. Yet according to Artificial Analysis, an average task in the Intelligence Index costs $2.29 with Sonnet 5, versus about $1.97 with Opus 4.8. At the maximum performance setting, Sonnet 5 burns through about 40 percent more output tokens per task than Sonnet 4.6. In agent-based benchmarks, it runs about three times as many agent loops as its predecessor, and Sonnet 4.6 cost about $1.20 per task. On CritPt, a frontier physics reasoning test, Sonnet 5 scored 17 percent, 14 points above its predecessor but below other models. Anthropic has done this before; when Opus 4.7 launched, a new tokenizer chopped the same text into "approximately 30%" more tokens, and a community analysis of over 483 submissions found a 37.4 percent jump in tokens per request.