Model Usage breakdown

Model distribution

The Model Usage dashboard shows a breakdown of API calls by model:

• •Call count per model
• •Cost per model
• •Average tokens per call by model
• •Trend over time

Finding optimization opportunities

Look for high-cost models handling simple tasks. Common patterns:

• •Opus for summarization — often Sonnet handles this equally well at 80% less cost
• •Large models for classification — Haiku typically handles classification tasks with equivalent accuracy
• •Default model everywhere — teams that haven't configured per-task model selection

Weekly digest

Your weekly digest email highlights the top model-switching opportunities with dollar amounts attached.