Claude Sonnet 5.5 Cuts Tokens and Tool Calls vs Sonnet 5

Anthropic released Claude Sonnet 5.5, priced at $2 input / $10 output with $2.50 for cache writes and $0.20 for cache reads. Anthropic states the model uses on average 30% fewer tokens than Sonnet 5. In practice, CodeRabbit's code-review pipeline found Sonnet 5.5's API calls cost about 40% of Sonnet 5's for equivalent reviews, a roughly 60% saving per review, attributing the gap to Sonnet 5's habit of reading the same files repeatedly and writing long deliberations. Enabling thinking mode added about 15% to cost but yielded one additional bug catch and slightly higher precision. In agentic testing, Sonnet 5.5 fixed a multi-bug Python project in three tool calls where Sonnet 5 required twelve. In a single informal coding session, Sonnet 5.5 completed a 3D brick-model designer app in about 29 minutes versus Opus 5.5's roughly 45 minutes, with outputs described as close to identical and Opus 5.5 showing slightly higher fidelity. The author of one roundup recommends Opus 5.5 over Sonnet 5.5 for most use cases despite the price difference.