Modelpublic

BattleGhost/Qwen3.8-Flash-Next-UNCENSORED-BF16-GGUF

sourceHugging Faceotherupdated 28d agoView on Hugging Face

0likes1.1kdownloads

Model Card

Qwen3.8-Flash-Next-UNCENSORED BF16 GGUF

BF16 GGUF conversion of `dealignai/Qwen3.8-Flash-Next-UNCENSORED-FP8`.

This repository contains the exact modified dealignai UNCENSORED checkpoint converted to GGUF. It is not derived from the independently modified OrcaRouter model.

Files

The model is split into eight GGUF files totaling 354,029,930,240 bytes (329.716 GiB). Start with Qwen3.8-Flash-Next-UNCENSORED-BF16-00001-of-00008.gguf; compatible llama.cpp tools discover the other splits automatically.

Conversion details

—Source revision: 2d9a479fd8643c394f1602b4ef1c8757dda4c518
—llama.cpp revision: c841aeeb8bb2fe417038dadfa9b007cf1a9ef950 (build b10687)
—GGUF architecture: qwen4exp
—Tensor count: 1,224
—Metadata fields: 67 in the BF16 split set
—Output type: BF16

The checkpoint stores the PLE n-gram FP8 shards with one shared ngram_embedding.weight_scale. The conversion applied that scale while materializing the BF16 PLE table and omitted the auxiliary scale tensor from the GGUF output. No refusal-removal or other behavioral modification was performed during conversion; the source checkpoint's modified weights were preserved.

Compatibility

Use a recent llama.cpp build containing Qwen3.8 / qwen4exp support. The relevant upstream implementation was merged in ggml-org/llama.cpp#27742.

License and safety

This is a format conversion of the dealignai checkpoint and remains governed by the Qwen Community License 1.0. Review the source model's license, model card, and disclaimer before use.

The source model has safety refusals removed. Users are responsible for complying with applicable laws and platform policies.