I’m using assistant api. And GPT 4o mini is really terrible in function calling. It both slow and not accurate. GPT 3.5 is on the other hand still faster and more accurate in function calling.
2 Likes
Completely agree. Run multiple tests and 3.5 just handles much better on average.
And 3.5 is much faster. I’m averaging 20% faster on my runs.
I don’t think OpenAI should deprecate this model. 4o hasn’t shown that is can actually replace it yet.