---

license: other

license_name: qwen-research

license_link: LICENSE

base_model:

• pottokao/Qwen-Image-2.1-PE-T2I-Heretic

tags:

• gguf
• quantized
• llama-cpp
• heretic
• abliterated
• prompt-rewriting
• qwen-image
• comfyui

---

Qwen-Image-2.1-PE-T2I — Heretic — GGUF

> Not affiliated with, or endorsed by, Alibaba / Qwen. Community derivative of

> Qwen/Qwen-Image-2.1-PE-T2I,

> redistributed under the Qwen Research License (copy included as LICENSE, per §3.a).

> Non-commercial use only; commercial use needs a separate licence from Qwen.

Q4_K_M build of

pottokao/Qwen-Image-2.1-PE-T2I-Heretic

— the Qwen-Image-2.1 prompt rewriter with refusal behaviour removed

(3/100 refusals @ KL 0.036, from 98/100 — best of a 1,600-trial search; dominates the earlier 7/100 @ 0.039, tag v1-trial384).

> system_prompt.txt is included and the model is useless without it. That 10 KB

> document is the output contract — structure, register, length, and the JSON shape.

> Load it as the system message.

Made with a simple node

The sample / showcase images in this repo were generated with the technique behind ComfyUI-QwenImage-PhotoStyles — a small ComfyUI node (17 photographic styles; one short prompt becomes a full styled prompt via the PE-T2I rewriter). If you're interested in how the images are made, the node and its recipe are there.

What this is for

A short request in any language goes in; one JSON line comes out with a detailed

English image prompt and a recommended aspect ratio:

{"rewritten_prompt": "...", "wh_ratio": "3:2"}

It is an LLM, not a text encoder. In ComfyUI it belongs in an LLM/GGUF node that

rewrites your prompt before it reaches the DiT — it does not load into

CLIPLoader. For the Qwen-Image-2.1 text encoder, see

…-Text-Encoder-Heretic.

Available quantizations

| File | Size | Notes |

|---|---|---|

| pe_t2i_heretic-Q4_K_M.gguf | 5.49 GB | smallest; token_embd kept at Q6_K |

| pe_t2i_heretic-Q6_K.gguf | 7.61 GB | higher quality; token_embd kept at Q8_0 |

| pe_t2i_heretic-Q8_0.gguf | 9.53 GB | near-lossless |

All load the same way (llama-server -m <file>); pick by available VRAM.

One deviation from stock Q4_K_M

token_embd is kept at Q6_K instead of Q4_K. Everything else is standard Q4_K_M

(llama.cpp already promotes output to Q6_K on its own).

The reason: every official NVIDIA NVFP4 checkpoint for this model family leaves the

token embedding untouched — across Qwen3-8B, Qwen3-32B, Qwen3-30B-A3B,

Qwen3.5-122B-A10B and Qwen3.5-397B-A17B, and also in the one official

mixed-precision recipe (Qwen3.6-27B), which quantizes lm_head to 4-bit but still

does not touch the embedding. Stock Q4_K_M does compress it. Raising it back costs

0.25 GB and cannot be worse than leaving it at Q4_K.

| | stock Q4_K_M | this build |

|---|---|---|

| token_embd | Q4_K (0.53 GB) | Q6_K (0.78 GB) |

| output | Q6_K | Q6_K |

| everything else | Q4_K / Q6_K | unchanged |

| total | 5.24 GB | 5.49 GB |

Run it

llama-server -m pe_t2i_heretic-Q4_K_M.gguf --host 0.0.0.0 --port 8080 \
             -ngl 99 -c 16384 --jinja -a pe-t2i

Then POST to /v1/chat/completions with system_prompt.txt as the system message.

Loads in a few seconds; fits on an 8 GB card.

Conversion

python convert_hf_to_gguf.py <bf16 model> --outfile pe_bf16.gguf --outtype bf16 --no-mtp
llama-quantize --token-embedding-type q6_K --output-tensor-type q6_K \
               pe_bf16.gguf pe_t2i_heretic-Q4_K_M.gguf Q4_K_M

--no-mtp is required. This model's config.json carries

text_config.mtp_num_hidden_layers with the value null. Without the flag the

converter does int += None and dies with a TypeError. The key existing with a

null value is the trap — .get(key, 0) returns None, not the default.

What was checked, and what wasn't

Checked — it still does its job. Four requests (three Chinese, one English)

through llama-server: 4/4 produced strictly valid JSON with exactly the

rewritten_prompt / wh_ratio keys, 2053–3351 characters. The bf16 source produces

2450–3320 on the same style of input, so the length discipline survives quantization.

Not checked — a clean perplexity comparison. I ran one and it returned

self-contradictory numbers (the unquantized bf16 GGUF scored worse than its own

quantizations, which is not physically sensible), so I am not reporting it. Take the

functional check above as the evidence, and nothing more.

Tried and not shipped — an imatrix build. I generated an importance matrix from

in-domain data and quantized with it. It passed the same functional check, but the

calibration corpus was badly designed: the 10 KB system prompt appears in every sample,

so ~88% of the 341 KB corpus was one repeated document, which would dominate the

importance statistics. Rather than ship a build I can't justify, it's left out. A

properly diversified corpus would be the way to revisit this.

🖼️ Showcase

<table>

<tr><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/01.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/01.jpg" width="100%"></a><br>_{Warm Documentary}</td><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/02.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/02.jpg" width="100%"></a><br>_{Candid B&W Street}</td><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/04.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/04.jpg" width="100%"></a><br>_{Vivid Everyday Color}</td></tr>

<tr><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/05.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/05.jpg" width="100%"></a><br>_{Black Fury}</td><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/06.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/06.jpg" width="100%"></a><br>_{Geometry of Light}</td><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/07.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/07.jpg" width="100%"></a><br>_{Frozen Film Still}</td></tr>

<tr><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/08.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/08.jpg" width="100%"></a><br>_{Deadpan Large-Format}</td><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/09.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/09.jpg" width="100%"></a><br>_{White Leap}</td><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/10.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/10.jpg" width="100%"></a><br>_{Corner Elegance}</td></tr>

<tr><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/11.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/11.jpg" width="100%"></a><br>_{Raw B&W Fashion}</td><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/12.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/12.jpg" width="100%"></a><br>_{Color Behind Glass}</td><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/13.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/13.jpg" width="100%"></a><br>_{High-Contrast Reportage}</td></tr>

<tr><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/14.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/14.jpg" width="100%"></a><br>_{Lyrical Landscape}</td><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/15.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/15.jpg" width="100%"></a><br>_{Hyper Chic East}</td><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/16.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/16.jpg" width="100%"></a><br>_{Street in the Mirror}</td></tr>

<tr><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/18.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/18.jpg" width="100%"></a><br>_{Plain Portrait}</td><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/19.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/19.jpg" width="100%"></a><br>_{Melancholy B&W}</td><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/20.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/20.jpg" width="100%"></a><br>_{Fairytale Giants}</td></tr>

<tr><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/21.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/21.jpg" width="100%"></a><br>_{Cold Power}</td><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/22.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/22.jpg" width="100%"></a><br>_{Cold Power}</td><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/23.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/23.jpg" width="100%"></a><br>_{Hyper Chic East}</td></tr>

<tr><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/24.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/24.jpg" width="100%"></a><br>_{Fairytale Giants}</td><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/25.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/25.jpg" width="100%"></a><br>_{Hyper Chic East}</td><td width="33%"><a href="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/blob/main/showcase_assets/26.jpg"><img src="https://huggingface.co/pottokao/Qwen-Image-2.1-DiT-GGUF/resolve/main/showcase_assets/26.jpg" width="100%"></a><br>_{White Leap}</td></tr>

</table>

This is what Q4_K_M across the board looks like — every image above was made with:

| Part | File |

|---|---|

| PE-T2I rewriter | pe_t2i_heretic-Q4_K_M.gguf |

| Text encoder | qwen3vl_8b_heretic-Q4_K_M.gguf + f16 mmproj |

| DiT | qwen_image_2.1-Q4_K_M.gguf |

| VAE | qwen_image_2.1_vae_bf16.safetensors (official) |

_{AI-generated. Style homage only — not affiliated with or endorsed by the photographers.}

Showcase

Outputs from the full pipeline this rewriter drives (rewriter → Qwen-Image-2.1

text encoder → DiT) — the image engine behind a "photography masters" system that

originally ran on DiffusionGemma and has since been moved onto this pipeline. Each

was generated from a one-line prompt; the rewriter expanded it into the scene,

palette and framing below. One seed per cell, no retouching.

<table>

<tr>

<td width="50%"><img src="showcase_assets/black_fury_night_dog.png" width="100%"></td>

<td width="50%"><img src="showcase_assets/black_fury_street_dog.png" width="100%"></td>

</tr>

<tr>

<td width="50%"><img src="showcase_assets/deadpan_portrait_pair.png" width="100%"></td>

<td width="50%"><img src="showcase_assets/geometry_of_light_stairs.png" width="100%"></td>

</tr>

<tr>

<td width="50%"><img src="showcase_assets/color_behind_glass_rain.png" width="100%"></td>

<td width="50%"><img src="showcase_assets/surreal_bw_street.png" width="100%"></td>

</tr>

<tr>

<td width="50%"><img src="showcase_assets/soft_everyday_light_child.png" width="100%"></td>

<td width="50%"><img src="showcase_assets/soft_everyday_light_harbour.png" width="100%"></td>

</tr>

<tr>

<td width="50%"><img src="showcase_assets/warm_documentary_portrait.png" width="100%"></td>

<td width="50%"><img src="showcase_assets/pale_light_glass.png" width="100%"></td>

</tr>

<tr>

<td width="50%"><img src="showcase_assets/vivid_everyday_diner.png" width="100%"></td>

<td width="50%"><img src="showcase_assets/vivid_garden_goldfish.png" width="100%"></td>

</tr>

<tr>

<td width="50%"><img src="showcase_assets/bold_minimal_flower.png" width="100%"></td>

<td width="50%"><img src="showcase_assets/soft_color_slip.png" width="100%"></td>

</tr>

<tr>

<td width="50%"><img src="showcase_assets/private_diary_flowers.png" width="100%"></td>

<td width="50%"><img src="showcase_assets/private_diary_portrait.png" width="100%"></td>

</tr>

<tr>

<td width="50%"><img src="showcase_assets/private_diary_street.png" width="100%"></td>

<td width="50%"><img src="showcase_assets/lyrical_landscape.png" width="100%"></td>

</tr>

<tr>

<td width="50%"><img src="showcase_assets/nocturnal_stage.png" width="100%"></td>

<td width="50%"><img src="showcase_assets/deadpan_industrial_tower.png" width="100%"></td>

</tr>

<tr>

<td width="50%"><img src="showcase_assets/minimal_seascape.png" width="100%"></td>

<td width="50%"><img src="showcase_assets/large_format_containers.png" width="100%"></td>

</tr>

<tr>

<td width="50%"><img src="showcase_assets/large_format_retail.png" width="100%"></td>

</tr>

</table>

_{AI-generated. Style homage only — not affiliated with or endorsed by the photographers.}

Other builds

| Repo | Format | Runtime |

|---|---|---|

| …-PE-T2I-Heretic | bf16 safetensors | transformers |

| this repo | GGUF Q4_K_M | llama.cpp / ComfyUI LLM node |

| …-PE-T2I-Heretic-NVFP4 | NVFP4 mixed precision | vLLM |

Results, the Pareto front of other operating points, and the measurement caveats are in

the base repo.