Post #2031716

axel. No one wants to use 4GB LLMs, though. Even the 20GB version of Gemma is frustratingly “stupid” and “insubordinate”. A 4GB model is going be impossibly unreliable. There’s no value in using these as a distributed data centre.