This developer routes 95% of tasks to DeepSeek and saves frontier models for the hardest 5%

mariofilhoml · x · 2026-10-24

Developer mariofilhoml shares his model-routing setup: DeepSeek V4.1 Flash 1017 handles 95% of tasks, while GPT 6.1 Sol is reserved for the 5% needing more intelligence — like exploratory data analysis where he refines hypotheses ('why is this bad?', 'why did this result change?') and wants answers probed through many nooks and crannies.

Related event: Developer's Pareto-optimal setup: DeepSeek for 95% of tasks, GPT for the rest(2 posts)→

Original post →

More from coding & agent

coding & agent channel →