>> when do you typically run different models? Do you put the same tasks to multiple models in parallel and compare the results? Or do you have a heuristic that model A is best for certain types of tasks and model B for others, etc?

I am not sophisticated enough here. :-) I have just been following on experience and recommendations on HN.

>> I have found that recently I seem to just use deepseek for almost everything. It is just good enough and much much cheaper than everything else.

Will try it soon. :-)

Thanks.