I feel that the issue of GPT being lazy is the most serious problem with current models (whether it’s Think or not, it will be lazy, Pro sometimes doesn’t output anything at all, automatically truncating). I hope Gemini 3 Pro doesn’t have similar problems.