Comment by p1esk

How do these small models compare to gpt4 for coding and technical questions?

I noticed that gpt3.5 is practically useless to me (either wrong or too generic), while gpt4 provides a decent answer 80% of the time.