I tried mudler's apex quant for gemma4 26b a4b and it was amazing! I got 38tps at 90.000 context with no loop and suprisingly no quality degradation. I used mudler/gemma-4-26B-A4B-it-APEX-GGUF / APEX-I-Compact (15gb) on my RX 9060 XT 16 GB with llama.cpp Vulkan. For comperison, my previous quant gem
