I'm slowly finding my way back to testing local LLMs (out of fear of price spikes). Currently positively surprised by OpenVINO. But my private benchmarks indicate not to trust edge models too much.

One question I'm asking myself is: can LLMs be bored?