I just spend a bit of time with o3-mini and it’s bad. I mean I’m trying to like it, it’s fast and all, but 2 things.
For planning, R1 is still the best, you can see the though process and you see things in there that sometime don’t even come out in the final answer. It will mention a tool like credo for example, and it won’t mention it in the output. I sometime read only what is in the think bracket.
For code gen, at least for Elixir, Sonnet 3.5 is still the king. o3-mini seem lazy (give you incomplete code) which I haven’t seen for a long time, there is a GPT-4 turbo vibe to it. Code quality itself, Sonnet is still far above.
Maybe 1 prompt app dev, o3 is better, I don’t know, not really interested in that use case.
What is your thoughts?