As the title says—for example, Gemini is actually very powerful! It’s just that developers have restricted it for cost reduction and efficiency gains, that’s all!
1 Like
The prompts definitely differ.
For example, the web version comes with built-in prompts. Today some guy was using AI Studio and said the latter doesn’t have default prompts, so it’s easier to use.
I actually think it’s a bit early to talk about “cost reduction and efficiency gains” right now,
at least judging from what I’ve seen so far, it hasn’t reached that level yet.
Everyone has also seen how full-size hard-run performs;
you can only say it’s barely satisfactory, and it’s still some distance away from being stably integrated into a workflow.
One relatively decent thing about “Gulu Gulu” is that
if it’s going to dumb things down for you, it will usually say outright that it’s because the quota has been reached,
at least it’s not pretending.
As for Gemini, based on what I’ve seen so far,
it’s probably about at that level—nothing too surprising.
On the contrary, there are some that are labeled 5.2,
yet tell the outside world they’re 4o—
I won’t expand on that 
1 Like