>>542301315
I remember it being pretty impressive. Pic related is the 13B model (which I enjoyed quite a bit at the time), and it got shit done.

As to whether botmakes should account for specific models? I would say no. There's such a variance in prompt style, frontend settings, bot defs, and more that trying to nail down a reliable standard for what constitutes a "good" bot design sounds like a fool's errand. At best, I would say you should note which model worked with your creation if someone wants to attempt to emulate your experience.

I dream of a future where everything is streamlined to the point that the bot has the necessary settings for optimal outputs hard-coded into it. You can choose to use them or not, but if you do, all your settings in ST (or whatever the future frontend is) automatically fix themselves in the background. This way, botmakies have no excuse for outputs not being consistent across the board for the {{user}}.