Qwen3.8 Flash AP Quants

Excerpt

Quite surprised to be beating other high quality quants. It took a lot of benchmarking to get here and we are quite pleased with these, hope they are useful to the community. It required a modified way of measuring KLD with a new dataset, since the NGRAM got in the way by remembering basically all of wikipedia. We tried to not only go for high precision, but also keep prefill performance in mind. Full model card here https://huggingface.co/agentionai/Qwen3.8-Flash-Next-AP-GGUF Let us know if the