brandon galang ▲
if you haven't given the new Gemini 3.7 Flash a fair shake, you're seriously missing out when you factor in token throughput, 3.7 flash is actually in a league of its own. it's seriously not even close coding benchmarks will put its total task time near frontier models like 5.6 sol, but in the planning/discussion stage gemini 3.7 flash is straight up 2-6x faster than comparable options. speed at this stage keeps you in flow and leads to better alignment with your agents and better execution gemini 3.7 as your "front door" model to clarify your scope with fable 5/gpt 5.6 sol as an advisor is a really ergonomic and effective setup, it's what i'm currently using
