license: apache-2.0
base_model: DavidAU/Qwen3.8-27B-Ultimate-Heretic-Uncensored
pipeline_tag: text-generation
library_name: gguf
description: "Seven NVFP4 GGUF tiers plus vision projector for DavidAU's Ultimate-Heretic Turbo-Brilliance-50x-AGI-NEO tune of Qwen3.8-27B: a byte-identical 448-tensor NVFP4 backbone across all tiers with a per-tier precision ladder for lm_head, token embedding, and MTP head. MTP speculative decoding baked into every file. Blackwell sm_120."
tags:

• gguf
• nvfp4
• qwen3.8
• blackwell
• mtp
• speculative-decoding
• ultimate-heretic
• llama.cpp
  language:
• en
• multilingual

Qwen3.8-27B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-NVFP4-MTP-GGUF

A family of seven GGUF files of DavidAU/Qwen3.8-27B-Ultimate-Heretic-Uncensored, DavidAU's Ultimate-Heretic Turbo-Brilliance-50x-AGI-NEO tune of Qwen3.8-27B: a 27B dense hybrid model (Gated DeltaNet plus Gated Attention every fourth layer, embedded MTP speculative head). The MTP head is baked into every file; no separate drafter is needed (--spec-type draft-mtp).

HF repo: esatapedico/Qwen3.8-27B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-NVFP4-MTP-GGUF

The classic 256-tensor NVFP4 checkpoint this family was built from lives in esatapedico/Qwen3.8-27B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-NVFP4 with its own card.

The 400-tensor W4A16 serving checkpoint lives in esatapedico/Qwen3.8-27B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-NVFP4-W4A16-T400 with its own card.

My part here is only the numerics: I converted the NVFP4 checkpoint to GGUF and built a size and precision ladder for the tensors that most affect output quality and decode speed. All credit for the model itself belongs upstream (full chain below).

Follow along & support

I post updates on new conversions, benchmarks, and what I'm working on over on my site. Follow along there to keep up with new releases and the work in progress. If you'd like to support more of it, you will find the options there. I do this on consumer hardware and like seeing how far it goes. More is on the way.

🌐 esatapedico.com. Updates, work-in-progress, and ways to support.

The seven files

Every tier carries the same byte-identical 448-tensor NVFP4 backbone (per-tensor SHA-256 verified) and differs only in the precision of the LM head, token embedding, and MTP draft head:

Picking a tier: MID-HIGH is the highest-precision compact option (all three head groups at Q8_0). COMPACT-LOW sits between VERY-LOW and LOW with a stronger LM head than VERY-LOW (Q4_K versus Q3_K). LOW and VERY-LOW trade some head precision for less VRAM. HIGH and VERY-HIGH restore BF16 heads where VRAM allows.

Tensor layout

Each tier holds 1,122 tensors. The shared 448-tensor NVFP4 backbone is byte-identical across all seven tiers (per-tensor SHA-256 verified): 256 tensors quantized in the NVFP4 conversion pass plus 192 normalized to NVFP4 when the tiers were built. The 48 ssm_alpha vectors stay F32.

Vision

Pair any tier with mmproj-BF16.gguf via --mmproj. Vision serving validation is pending.

Measurements

Perplexity on a source-diverse corpus (lower is better):

Perplexity stays within about 1.3% of the unquantized 3.2445 figure across the whole ladder, so the head ladder costs almost nothing even at the smallest tier.

slimmed-full MMLU-3000 classic 85.73% vs ninfer 82.88% (same subset_sha bf45e6bd, SE ~0.69pp)

Usage

llama.cpp / llama-server

llama-server \
  --model Qwen3.8-27B-Ultimate-Heretic-Turbo-Brilliance-50x-AGI-NEO-NVFP4-MTP-MEDIUM.gguf \
  --ctx-size 131072 \
  --flash-attn on \
  --spec-type draft-mtp \
  --temp 0.7 --top-p 0.95 --top-k 20

• Requires a recent llama.cpp with NVFP4 (GGML type 40) CUDA kernels and Blackwell support.
• Requires the draft-mtp spec path for the embedded MTP head.
• For vision input, add --mmproj mmproj-BF16.gguf. Vision serving validation is pending.

How this was made

1. Quantized with a datafree NVFP4A16 recipe: targets Linear, ignore list lm_head, visual tower, MTP head, conv1d, in_proj_a, in_proj_b. Sequential run, CPU only, no calibration corpus.
2. Packed 400 NVFP4 tensors. The 15 MTP tensors were restored to BF16 after quantization.
3. Converted to GGUF with convert_hf_to_gguf.py --outtype auto, keeping native NVFP4 tensors intact.
4. Built each tier with llama-quantize --tensor-type-file over the shared backbone.
5. Verified: per-tier NVFP4 census (448 NVFP4 tensors each) and per-tensor SHA-256 backbone identity across all tiers.

Attribution & provenance

This is a derivative work built entirely from existing Apache-2.0 artifacts. Nothing here was trained or fine-tuned. This work covers the numeric conversion only and makes no ownership claims over the model. Credit belongs to:

1. Alibaba / Qwen team for the base model family (Apache-2.0): the 27B dense hybrid architecture this tune builds on.
2. DavidAU for the tune itself, Qwen3.8-27B-Ultimate-Heretic-Uncensored (Apache-2.0), including the Turbo-Brilliance-50x-AGI-NEO system this family is named for.
3. This repo's author for the NVFP4 conversion, the GGUF conversion, and the tier ladder only.

License

Apache-2.0, identical to every upstream artifact. The base model license governs. GGUF conversion and quantization are transformations, not new training. When redistributing, please retain attribution to Qwen (Alibaba) and DavidAU as above.

"Qwen" is a trademark of Alibaba. Trademarks are used here only to identify upstream models. This repository is not affiliated with, sponsored by, or endorsed by Alibaba or DavidAU.

Card written with AI assistance.