Qwen3.8-Flash-Next EXL3 3.05 bpw for mlx-serve (moved)
Moved to OpensourceWTF/Qwen3.8-Flash-Next-exl3, branch 3.05bpw.
Every file this repo held at commit 71797b00, except ngram_embedding.safetensors, is on that branch with the same
bytes, each checked against this repo's sha256s. Make the n-gram table from turboderp's file on his branch
3.05bpw_h5_ng5 (sha256 32006bf33c1e…) with the converter's header step, as the branch's card says. This repo's history was squashed to free storage.
The quantization is turboderp's: turboderp/Qwen3.8-Flash-Next-exl3 (branch 3.05bpw_h5_ng5), made with
exllamav3. Every branch, its data and its charts:
OpensourceWTF/Qwen3.8-Flash-Next-exl3.