It's not for general use, but sometimes when I get stuck or something goes wrong while setting up a large model on AI MAX,
I solve it by mobilizing the API. I use DeepSeek or GLM flash series, which are relatively inexpensive on OpenRouter.
But no matter how cheap it is, after solving a few things, $2-3 flies away -_-:
And even though the results are the same, it seems like the API is consuming a bit too much.
When I had my source code analyzed last time, DS V4.1 Flash cost $2 and was just thinking, so I turned it off.
SNS also says to use a local model of about 10B, asking why API is much cheaper,
but in the end, I realized that the API doesn't cost much.
