Text this: Yet another approximation of human semantic judgments using LLMs... but with quantized local models on novel data