Model Router — which AI model do you actually need?

Defaulting to the most powerful model for every task is the biggest single source of wasted tokens, cost, and hidden water. Pick one of five task types and rate three factors — output precision, reasoning complexity, context length — to get a model-tier recommendation. The reusable rule: precision + reasoning + context + task bias → score ≤7 fits Haiku, 8–12 Sonnet, 13–15 Opus, 16+ Fable 5. Routing a fitting task to Haiku instead of Sonnet saves about 88% water (an illustrative per-tier ratio, not a measurement — see methodology) and 67% cost (input list prices as of Jul 2026).

Part of the Sustainable AI Toolkit: all tools · methodology & sources · Prompt Efficiency · Eval Bench · Water Footprint

This page requires JavaScript to run the interactive router.