meshllm/DeepSeek-V4-Flash-UD-Q4_K_XL-layers
Quantization (fewer layers) vs unsloth/DeepSeek-V4-Flash-GGUF. meshllm/DeepSeek-V4-Flash-UD-Q4_K_XL-layers matches part of unsloth/DeepSeek-V4-Flash-GGUF (27 of its 1328 tensors; the rest of unsloth/DeepSeek-V4-Flash-GGUF has no counterpart). On the shared part: meshllm/DeepSeek-V4-Flash-UD-Q4_K_XL-layers is a GGUF quantization (MXFP4/Q8_0) of the same weights: 25 tensors match exactly and 2 within quantization error.
Tensors changed: other 0/3, Attention 0/9, Norms 0/4, MLP 0/11