Change default model to gpt-5

[werdnum]

I found on a new machine that llm was giving me rubbish results. It turned out the reason for this was because the default model is gpt-4o-mini. This is a pretty old model now! Should set it to gpt-5 (probably).

[mqudsi]

gpt-5 is dog slow. fyi you can always set your own default.

vs

It's over 10x slower than the full 4.1 model (E2E latency)