Token optimization strategies
Hi all
Got Copilot Business at work, came across a recommendation to use GPT 5.6 Luna as an explore agent instead of default Gemini flash 3.5 or Haiku 4.5 whatever it was using.
Any more tips and tricks like that? I am using Ubuntu on the laptop btw. Plus how's the Plan+execute pattern looking like using Opus /Sonnet to plan and then maybe use MAI Flash 1.1 to execute? MAI 1.1 looks quite inexpensive.