By skyforbes But as I do more and more detailed, complex, intricate bespoke work, the cracks are showing fast.
GPT 5.1 Thinking has been a master of consistency and meticulousness for me. I had seen no AI before it like this. But Gemini, while it is super good at rapidly shooting out big, impressive things, it is nowhere nearly as good as GPT Thinking in consistently capturing every detail of a dataset, nuance, instruction, and delving in with consistent effort and care – if that makes sense.
It's also what makes GPT Thinking annoying sometimes – it takes forever to do something that Gemini 3 does in 1/5th to 1/10th the time sometimes. But then you realize just how much Gemini keeps missing and it almost feels hopeless to try to get it to work right. The more you guide it to fix one thing, it will stuff up or miss something else it had right before.
I find Gemini 3 Thinking does "beautiful" work; GPT 5.1 Thinking does meticulous work.
Gemini 3 is a talented passionate artist, GPT is a talented dedicated analyst.
I'm always more impressed by Gemini 3 Pro at first sight. I always end up trusting GPT 5.1 Thinking more with what matters.
(Note: I wouldn't trust GPT 5.1 non-thinking with a plastic cup.)