Gemma 4 and Qwen 3.6 with q8_0 and q4_0 KV cache: KL divergence results

r/LocalLLaMA
Open Source AI

Submitted by /u/oobabooga4 [link] [comments]