Qwen3.8-27B Uncensored Heretic — MTP, sized for 12 / 16 / 24 GB GPUs

GGUF quants of llmfan46/Qwen3.8-27B-Uncensored-Heretic-Native-MTP-Preserved, built so the model and its native MTP head fit fully in VRAM on common consumer cards. Each file applies Unsloth's per-tensor "UD" recipe for the base Qwen3.8-27B, copied tensor by tensor, to llmfan46's weights. The MTP block (blk.64) stays at q6_K / q8_0 in every tier.

Every claim below was measured; the methodology is at the end.

Files

File	Size	Target card	KLD vs Q8_0 (prose / code)
`Qwen3.8-27B-Uncensored-Heretic-MTP-UD-IQ3_XXS.gguf`	10.18 GiB	12 GB	0.090 / 0.059
`Qwen3.8-27B-Uncensored-Heretic-MTP-UD-IQ4_XS.gguf`	13.27 GiB	16 GB	0.027 / 0.019
`Qwen3.8-27B-Uncensored-Heretic-MTP-UD-Q5_K_XL.gguf`	19.44 GiB	24 GB	0.0045 / 0.0037

Quality against the other published quants of the same weights

Mean KL divergence against a Q8_0 reference of the same BF16 (lower is better), with the share of tokens whose top-1 prediction matches the reference.

Tier	Quant	GiB	KLD prose	KLD code	Same top-1 (prose / code)
24 GB	this repo UD-Q5_K_XL	19.44	0.0045	0.0037	97.2% / 98.5%
16 GB	mradermacher i1-IQ4_XS ¹	14.26	0.0202	0.0149	94.3% / 96.8%
16 GB	this repo UD-IQ4_XS	13.27	0.0268	0.0192	93.2% / 96.5%
16 GB	llmfan46 Q3_K_M	13.48	0.0639	0.0456	89.3% / 95.0%
16 GB	mradermacher i1-IQ3_M	11.89	0.0649	0.0465	89.8% / 94.9%
12 GB	this repo UD-IQ3_XXS	10.18	0.0904	0.0587	87.4% / 94.5%
12 GB	mradermacher i1-Q2_K	10.12	0.1551	0.1017	82.9% / 92.6%

¹ Better than UD-IQ4_XS, but at 14.26 GiB it does not leave room for MTP plus a useful context on a 16 GB card.

16 GB: 2.4× lower KLD than the two quants that fit with MTP (llmfan46 Q3_K_M, which is also 0.2 GiB larger, and mradermacher i1-IQ3_M).
12 GB: 1.7× lower KLD than mradermacher i1-Q2_K at the same size.
24 GB: effectively indistinguishable from Q8_0. No other 24 GB-class quant was measured, so no comparison is claimed.

Speed (RTX 4080 16 GB, measured)

llama.cpp b11457 (CUDA 13.4), MTP on, 2 draft tokens, ubatch 256, q4_0 KV cache, 32K context, a 15,216-token prompt (a C++ source file), 600-token answers, mean of 3 fixed seeds.

Quant	Code	Prose	MTP acceptance (code / prose)	Prompt eval
this repo UD-IQ4_XS	55 tok/s (49–62)	50 tok/s	83% / 58%	~1,360 tok/s
this repo UD-IQ3_XXS	77 tok/s	65 tok/s	83% / 61%	~1,440 tok/s
mradermacher i1-IQ3_M	70 tok/s	54 tok/s	83% / 54%	~1,480 tok/s
mradermacher i1-Q2_K	72 tok/s	59 tok/s	79% / 53%	~1,160 tok/s
llmfan46 Q3_K_M	42 tok/s (32–50)	41 tok/s	84% / 61%	~1,200 tok/s
UD-IQ4_XS without MTP (ubatch 512, unseeded)	29 tok/s	29 tok/s	—	~1,500 tok/s

MTP roughly doubles decode speed at this depth. Two draft tokens beat three on the same seeds (68 vs 56 tok/s code, 55 vs 36 prose, at 40K), because fewer drafts get rejected.
Keeping the MTP head at q6_K changes acceptance very little (83% vs 83% on code, 58% vs 54% on prose against a quant with the head at IQ3_S). The quality gain of these files comes from the trunk, not the head.
The UD-IQ3_XXS numbers come from a 16 GB card with headroom. A 12 GB card has less memory bandwidth and will be slower.

Recommended settings

16 GB — UD-IQ4_XS

llama-server -m Qwen3.8-27B-Uncensored-Heretic-MTP-UD-IQ4_XS.gguf -ngl 99 --flash-attn on \
  -c 32768 -ub 256 -ctk q4_0 -ctv q4_0 -np 1 \
  --spec-type draft-mtp --spec-draft-n-max 2

This sits at the edge of 16 GB. Measured on the same card on different days, 40K of context gave 68, 51 and 42 tok/s on code with the desktop holding ~1.2, ~1.4 and ~1.9 GB of VRAM. When the model spills, Windows moves part of it to system RAM ("shared GPU memory" in Task Manager) and decode drops. During one run, opening a Chrome window mid-generation dropped decode to 0.2 tok/s. If you run a desktop on the same GPU, use 32K or less and watch shared GPU memory: it should stay near its idle value. 48K with MTP produced the same tokens as 40K and was 22% slower, from spill alone.

12 GB — UD-IQ3_XXS

llama-server -m Qwen3.8-27B-Uncensored-Heretic-MTP-UD-IQ3_XXS.gguf -ngl 99 --flash-attn on \
  -c 16384 -ub 256 -ctk q4_0 -ctv q4_0 -np 1 \
  --spec-type draft-mtp --spec-draft-n-max 2

Not run on a 12 GB card. The context is computed from the buffers llama.cpp reported at 32K: weights 10,020 MiB, recurrent state 449 MiB, compute ~260 MiB, KV ~22 KB per token including the MTP draft cache, about 11.6 GB in total at 32K. On a headless 12 GB card ~32K should fit. With a desktop on the same card, start at 16K.

24 GB — UD-Q5_K_XL

Same flags with -c 65536. Also computed rather than measured: ~22.4 GB at 64K.

Caveats

Thinking. The chat template accepts reasoning_effort = none, low/minimal, medium (default) or high/xhigh, plus chat_template_kwargs.enable_thinking. Thinking stays on when tools are present unless auto_disable_thinking_with_tools is set.
Uncensored. Refusal removal is llmfan46's work (Heretic v2, MPOA); their card reports 3/100 refusals and KLD 0.0244 against the original Qwen3.8-27B. That was not re-measured here, and quantization can make behaviour near the old refusal boundary less stable. The model will comply with requests the original refuses; you are responsible for how you use it.
Vision. These files are text-only. llmfan46's mmproj was built for the same weights and should work for image input, but it was not tested with these quants.

How these were made

BF16 GGUF from llmfan46 (SHA-256 088e2801…5a062).
The per-tensor types of unsloth/Qwen3.8-27B-GGUF (UD-IQ3_XXS, UD-IQ4_XS, UD-Q5_K_XL) read from the GGUF headers. All 866 tensors match llmfan46's BF16 in name and shape.
llama-quantize --tensor-type-file with mradermacher's imatrix for these exact weights. Every recipe was dry-run and checked tensor by tensor before quantizing. Final sizes match Unsloth's files.

One gotcha if you do this yourself: --tensor-type-file patterns are std::regex_search patterns. A plain output.weight=q5_k line also matches every blk.N.attn_output.weight and silently re-types them. Anchor and escape each name: ^output\.weight$=q5_k.

Evaluation. Reference: Q8_0 from the same BF16. Text: wikitext-2 test and two llama.cpp source files (~260 KB). llama-perplexity --kl-divergence, context 1024, 40 chunks per text (~20K scored tokens each, standard error ±3%). Speed: llama-server as above, prompt from server-task.cpp. Scripts are in scripts/.

Credits

Qwen for Qwen3.8-27B.
llmfan46 for the Heretic weights with the MTP head preserved.
Unsloth for the dynamic per-tensor recipes.
mradermacher for the imatrix.
llama.cpp.

License: Apache-2.0, as the base model.

codavidgarcia/Qwen3.8-27B-Uncensored-Heretic-MTP-UD-GGUF