You can always claim a model is different or unique compared to GPT 4 in terms of size , 1 narrow benchmark etc

But to make blanket statements like far better than GPT-4 sounds like marketing gimmick .

Also specific examples where Gemini works better would have helped . I have played many models on various text, coding and reasoning tasks , nothing came close to gpt 4 in most scenarios

Have tried Gemini too few times . Few because if it doesn’t give any benefit I continue with GPT4 as my goal is getting it done with best model quickly