Erase Thinking portion of llm response after it's been completed

I want the ability to delete the Thinking portion of DeepSeeks responses and future reasoning models to save on context. As far as I understand it the whole chat is sent to the model as context each prompt, not only does the thinking segment take up a up to 1/3 of the context but the Thinking portion is often prone to confusion. By confusion I mean it’ll think about what I might mean and not mean, but answer mostly correctly. However future prompts mistake that portion of the previous prompt for my actual meaning and discard what I’ve explicitly noted.

Being able to toggle deleting thought process after a response has been fully generated would be valuable for both these reasons imo.