In my project, I’ve identified which models are best for different types of tasks, such as code writing, documentation, and refactoring. I want to be able to make a Plan and then choose a different model for each step. I do not think it is currently possible.
Hm.. my experience, i use for create plan using Model such as GPT-5 or Big Model, but when come to execution plan, use different model such as Grok Code, Haiku or Gemini 3 Flash.
Current workaround: You can use a stronger model (like GPT-5 or Opus) to create the plan, then manually switch to a faster or cheaper model (Grok Code, Haiku, Gemini Flash) before running Build. This helps you balance planning quality with execution speed and cost.
I’ll pass this to the team as a feature request. If you have specific use cases, like which models you want for which task types, share them and it’ll help the team prioritize.