Shows which models developers actually pair for planning versus execution, such as Opus 4.1 planning with Sonnet 4 acting. Gives a real-usage baseline for choosing a cost-aware model split.
Key takeaways · AI-distilled
In Cline's seven-day sample, published in October 2025, Gemini 2.5 Pro was the second most used plan modeA read-only agent mode that investigates and proposes a plan without editing anything, so you approve the approach before any code changes.Full definition → model at 15.3%, roughly a third of Claude Sonnet 4's 42.6% share.
Among sessions using different models for planning and acting, the Opus 4.1 plan and Sonnet 4 act pairing made up 25.3%. At the post's OpenRouter prices, Opus 4.1 ($15/$75 per million tokenThe chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.Full definition →) cost five times Sonnet 4 ($3/$15).
Cline said it expected Claude Sonnet 4.5, priced the same as Sonnet 4 at $3/$15 per million tokens, to replace Sonnet 4's usage in both Plan and Act modes over the following weeks. That was a forecast, not measured adoption.
Terms in this piece · Glossary
plan mode — A read-only agent mode that investigates and proposes a plan without editing anything, so you approve the approach before any code changes.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
