Bottom-line, up front:
Which AI model as of the current date, do you think is the best for coding/programming tasks?
Yes, this might be subjective but I would be glad, if you could also state why, you regard this as the best.
… and yes, I will use “AI” as the popular, generic term for all things LLM, ML etc.
The longer version:
We’re a small group of amateur programmers and the subject of pair programming has come up. So of us has tried it in a work setting with different results. However pair programming with an AI is something newer and while all of us has used AI to create boilerplate code etc. none of us, has really tried to use it as a “paired programmer”.
My own experience is with ChatGPT 4o (and a couple of the earlier models) and GitHub Copilot (the IDE integrated one).
Working primarily with C#, I am not impressed by Copilot. Both C# and Copilot is a developed and trained under the umbrella of Microsoft and still it doesn’t understand langugage features, which were released over a year ago, e.g. using primary constructors for classes. Also it seems stateless, so in the same thread, it will suggest bonkers stuff. I have more grievances about Copilot, but let us move on.
ChatGPT 4o is nice enough but needs a lot of instruction. However it seems way more stateful than Copilot, which is nice for a bit larger products. It feels a little bit like having an eager junior dev 
I have heard others talk about, that some of the other larger models is way better at producing code, so it might be a better match to use for pair programming.
What are your experiences. Also tips and hints are very welcome 