Interesting - which models specifically? I'd be interested in using mistral over deepseek if it was competitive (guess I need to go benchmark)
I use small, large, an medium-3.5 depending on the task
I use small, large, an medium-3.5 depending on the task