What TPS is too slow for you?

The discussion revolves around acceptable Tokens Per Second (TPS) for running local models. Some users are content with sub-30 TPS for high-quality models, while others prioritize speed, aiming for 100-150 TPS or more, even if it means compromising quality. The core question is to define the absolute minimum TPS that users would tolerate for their local models.

This discussion uniquely highlights the wide disparity in user expectations for local model performance, ranging from sub-30 TPS for quality to over 100-150 TPS for speed.