The 80% rule: when Haiku beats Opus

Default to the biggest model? That's expensive.

It's the path of least resistance: pick the most capable model, never think about it again. But capability you don't need is cost you don't need.

Where smaller models match

Our data shows that for these task types, mid-tier and fast models match flagship quality 80%+ of the time:

• -Summarization — Sonnet and Gemini Flash handle this as well as Opus
• -Code review — catching bugs and style issues doesn't need the biggest model
• -Classification — routing, labeling, and categorization are well-solved at every tier
• -Simple generation — boilerplate, templates, and standard patterns

Where flagship models matter

• -Complex multi-file reasoning — understanding interactions across a codebase
• -Architecture decisions — tradeoff analysis and system design
• -Novel problem solving — tasks with no clear pattern to follow
• -Long-context synthesis — pulling insights from 100k+ token contexts

The 80% rule in practice

Look at your last 100 API calls. Roughly 80% of them probably fall into the first category. If you're paying flagship prices for all of them, you're overspending by 3-5x on those calls.

The savings are immediate and the quality difference is undetectable for routine work.