Why proprietary weight models don’t make sense any more

I only have to say DeepSeek V4 Flash (and even worse case with the newer version), and you don't need the ultimate best model most of the time if you give the correct context you can get away most of the time with this cost efficient model.

Here is is some data:

So at least if you are using text only, DeepSeek V4-Flash 0731 is a pretty competitive option. Even though there are more capable models, this model is capable enough for the price.