imatrix GGUF quants Execution tips from my private experience: don't quantize context use top_p 0.9, top_k 20, temp 0.6, min_p 0.05
