I personally use GPT 4.1 in simple ask mode most recently. Fast and usually correct for quite complex function so OpenAI seems to be winning IMO.

All these "agentic" things make these models so confused that it almost never gives good results in my testing.