LM Studio users push for smarter per-model defaults

This is less a beginner mistake than a UX warning for the whole local LLM stack: one-size-fits-all defaults are breaking down as model families diverge on sampling and load behavior.

• –LM Studio’s docs already support per-model default load settings, which helps, but that is not the same as auto-detecting the right starter values for every downloaded model.
• –Its preset system can save inference parameters like temperature and top-p for reuse, yet load parameters are handled separately, which adds more mental overhead for casual users.
• –The bigger opportunity is a model registry or metadata layer that ships recommended sampler settings with each model so local inference feels more like install-and-go software than manual tuning.