AesSedai/Qwen3.8-Flash-Next-GGUF
1911k
Updates / Notes
- 09/02/2026: Added the PLEQ40 for IQ2S and IQ3_S
- 09/01/2026: The original quants all have Q80 for the engram embedding tensors, and I've uploaded three variants with Q40 for the embedded tensors. The only difference is the Q80 vs Q40 for those tensors. I'm leaving the original quants up for those who want to use the Q8_0 PLE's.
This repo contains specialized MoE-quants for Qwen/Qwen3.8-Flash-Next. The idea being that given the huge size of the FFN tensors compared to the rest of the tensors in the model, it should be possible to achieve a better quality while keeping the overall size of the entire model smaller compared to a similar naive quantization. To that end, the quantization type default is kept in high quality and the FFN UP + FFN GATE tensors are quanted down along with the FFN DOWN tensors.