Qwen3.6-27B KV quantization comparison: Q8 vs Q6 vs Q5

A Reddit user measured Kullback-Leibler divergence for KV cache quantization levels Q8, Q6, and Q5 on Qwen3.6-27B. Results show Q6 and Q5 introduce increasing divergence, with implications for long-context performance.