Sorry I am very new to this stuff.
Overwhelmed with choice 
I think I am in safe hands using the ones I hear about the most: GPT4o and Sonnet 3.7
However, I’ve heard Gemini (was it the 2.5 pro or exp-1206?) has a much much bigger context window, which supposedly works better as the project become closer to something serious…
As you can see, I am still finding my feet in this world and would love to know from someone with experience about the “huge” context window models - maybe the bottleneck is still with the accuracy/reasoning and the extra context on Google’s stuff isn’t useful yet?
My priority is high quality code that I can learn from and can somewhat trust while I am learning.
Yes I’ve been using Gemini 2.5 Pro and occasionally Gemini 2.5 Pro(MAX) for the more intensive requirements. Previously I primarily used Claude 3.7 but I’ve come to prefer Gemini and have been using it as my primary model.
One thing I should note, I noticed that once the project starts getting larger changing to a different providers model seems to cause confusion or a much different code style.
It creates inconsistent results, it may work but the coding style may be much different and there will be suggestions/requests to modify perfectly good code to conform to the new additions.
For example, starting with Gemini 2.5 pro and building a good working foundation then later switching the model to Claude 3.7. Trying this has not been ideal in my projects once they start growing in complexity.
But yes, Gemini 2.5 Pro and Max has been a great model for me.