With the new quota model, have you guys noticed a drop in performance when using credits?
What I’ve observed is that when I’m working within the baseline 5-hour quota (yes, I have the Ultra plan — though I’m canceling it for reasons I think many of you can guess), the model is fast and performs very intelligently. However, once that quota is exhausted and I start using credits, the model becomes slower and starts making more mistakes, almost as if it’s less capable or somehow “quantized.”
Has anyone else noticed the same behavior?