license: apache-2.0
base_model:
- AtomicChat/Qwen-Image-2.1-Turbo-Uncensored-GGUF
- Comfy-Org/Qwen3-VL
- Qwen/Qwen3-VL-8B-Instruct
base_model_relation: quantized
library_name: diffusion-single-file
pipeline_tag: text-to-image
tags: - comfyui
- qwen3-vl
- qwen-image-2.1
- text-encoder
- bf16
- int8
- convrot
- abliteration
- not-for-all-audiences
Qwen3-VL 8B Abliterated · INT8 ConvRot & BF16
A complete multimodal Qwen3-VL text encoder for Qwen Image 2.1 Turbo in ComfyUI, reconstructed from AtomicChat's BF16 GGUF release. Choose the 9.35 GB INT8 ConvRot checkpoint for reduced storage, or 17.53 GB BF16 for the original reconstructed precision.
Both include the entire BF16 vision tower and image-to-language mergers needed for image conditioning. No separate mmproj is required. The source modification affects the language encoder; the diffusion transformer and VAE are unchanged.
Choose your checkpoint
| Checkpoint | Download | Exact size | Precision |
|---|---|---|---|
| INT8 ConvRot | qwen3vl_8b_abliterated_int8_convrot.safetensors | 9,350,798,360 bytes · 9.35 GB | INT8 language matrices; BF16 vision and norms |
| BF16 | qwen3vl_8b_abliterated_bf16.safetensors | 17,534,334,976 bytes · 17.53 GB | Complete BF16 reconstruction |
The INT8 checkpoint matches the official Comfy-Org encoder's exact file size and tensor precision policy. It is 46.67% smaller than BF16. Each file is a complete encoder; download one for ordinary use.
Quick start in ComfyUI
- Download your chosen checkpoint into
ComfyUI/models/text_encoders/. - In CLIPLoader, select the checkpoint and set type to
qwen_image. - Connect the CLIP output to TextEncodeQwenImage21 in a Qwen Image 2.1 Turbo workflow.
- Use the matching diffusion model and VAE below. For image editing, also connect your reference image and VAE to TextEncodeQwenImage21, then use its latent output as the sampler's input.
CLI download, from the directory containing your ComfyUI installation:
hf download NidAll/Qwen3VL-8B-Abliterated-INT8-ConvRot \
qwen3vl_8b_abliterated_int8_convrot.safetensors \
--local-dir ComfyUI/models/text_encoders
To download BF16 instead, substitute qwen3vl_8b_abliterated_bf16.safetensors.
Required companion models
Download these separately from Comfy-Org/Qwen-Image-2.1:
| Component | Tested file | ComfyUI folder |
|---|---|---|
| Diffusion model | qwen_image_2.1_turbo_bf16.safetensors |
models/diffusion_models/ |
| VAE | qwen_image_2.1_vae_bf16.safetensors |
models/vae/ |
For the tested Turbo setup, use Euler, CFG 1, eight steps, with these exact sigmas:
1.0, 0.978453, 0.95418, 0.926626, 0.89508,
0.845148, 0.704534, 0.414568, 0.0
Image-editing tests used the same reference image across encoder variants. Use the reference image's dimensions and preserve the reference conditioning produced by the encoder node.
The tested ComfyUI revision is 0df64eb242b7c5759c3e86afd5d1846d923b1033, with comfy-kitchen==0.2.37. An older ComfyUI build may lack Qwen Image 2.1 support or the required inline INT8 ConvRot/embedding loader.
What is included
Language: all 36 transformer blocks, token embeddings, normalization tensors and the language output head. Exactly 73 tensors differ from the original BF16 reference: 36 attention-output projections, 36 MLP down-projections and the token embedding.
Vision: the complete 27-block tower, patch and positional embeddings, visual attention/MLPs, normalization, the main merger and three DeepStack mergers. All 351 final vision tensors remain BF16 and were verified numerically against the BF16 mmproj. The GGUF's two temporal patch-convolution slices were correctly reassembled.
The reconstructed checkpoint contains 750 BF16 tensors. GGUF BF16 bits were extracted directly, without an FP16 intermediate. GGUF FP32 tensors were reused in BF16 only after exact equality with the widened reference was established. All 677 unchanged reference tensors are preserved bit for bit.
INT8 ConvRot policy
| Setting | Value |
|---|---|
| Native storage format | int8_tensorwise |
| Quantized matrices | 254: 252 attention/MLP projections, token embeddings and lm_head |
| Transformer coverage | All language blocks 0–35 |
| Rotation | Regular orthogonal H4 ConvRot, group 256 |
| Scaling | FP32 per output row |
| Conversion arithmetic | FP32, bounded butterfly transform |
| BF16 preservation | Complete vision/mergers, norms, biases and remaining official BF16 tensors |
| Weight error gate | Relative L2 ≤ 0.012 for every selected matrix |
Conversion started from the validated BF16 checkpoint, using the ComfyUI Native Quantizer codec and a pinned official tensor-layout template. No activation calibration, learned rounding, rank reduction or double quantization was applied. Every selected matrix passed the gate; there are no BF16 boundary-block exclusions.
The final file has 1,258 stored tensors: 254 INT8 weights, 496 BF16 tensors, 254 FP32 scales and 254 inline quantization metadata tensors. 93.42% of model parameter elements are INT8.
Storage and execution
Stock ComfyUI CLIP encoding can retain INT8 storage while performing floating-point matrix multiplication after inverse rotation. The filename does not imply automatic INT8 GEMM acceleration. The quantized embeddings use native token-row dequantization with inverse ConvRot.
A separate test explicitly enabled native quantized operations and disabled forced floating casts in runtime test objects. It observed 504 CUDA INT8 linear calls per generation, plus native embedding dequantization and INT8 tensor-core GEMM kernels. That experimental compute mode is not automatically enabled by ordinary CLIPLoader use.
Validation results
Reconstruction passed complete tensor coverage, serialized identity, tokenizer and NaN/Inf checks. Source reconstruction errors were zero; all 151,669 token IDs and 151,387 BPE merges matched. Independent comfy-kitchen decoding of all INT8 matrices produced:
| Weight metric | Result |
|---|---|
| Mean per-matrix relative L2 error | 0.00887948 |
| Maximum per-matrix relative L2 error | 0.01020596 |
| Minimum per-matrix cosine similarity | 0.99994792 |
16 fresh INT8 generations—eight stock-compute and eight native-INT8-compute runs—were compared against eight unchanged BF16 baseline images. All used 1024-pixel output, seed 42, identical prompts/reference image, diffusion model, VAE and sampling settings. Every image edit exercised the vision tower and three DeepStack mergers; all conditioning, latent and decoded-image tensors were finite.
Matched image similarity results
SSIM measures image similarity, not prompt adherence or semantic correctness.
| Case | Stock INT8 SSIM | Native INT8 compute SSIM |
|---|---|---|
| Still life | 0.9900 | 0.9839 |
| Long spatial prompt | 0.7676 | 0.9087 |
| English typography | 0.9770 | 0.9914 |
| Chinese typography | 0.9825 | 0.9867 |
| Ordered objects | 0.9554 | 0.9569 |
| Object replacement | 0.9972 | 0.9967 |
| Style edit | 0.9983 | 0.9984 |
| Lighting/structure edit | 0.9981 | 0.9980 |
Manual inspection confirmed readable English/Chinese text, the requested object order and successful teapot-to-watering-can replacement in these examples. The library scene showed placement/detail differences. Watercolor and dusk changes were weak in BF16 and both INT8 modes: those cases passed execution but did not pass semantic transformation acceptance.
These measurements cover eight prompts at one seed. They do not establish universal quality equivalence, robust typography, broad style editing, multi-reference behavior, chat generation quality or refusal reduction. Weight error is not activation error. No portable latency or memory-capacity claim is made.
Sources and attribution
| Source | Pinned revision | Role |
|---|---|---|
| AtomicChat/Qwen-Image-2.1-Turbo-Uncensored-GGUF | e3b01e9bfacc373148bdb93824496e0c72643250 |
Modified BF16 language GGUF and BF16 mmproj |
| Comfy-Org/Qwen3-VL | 02f0d3eefb59528799653d16735818801177ef1f |
text_encoders/qwen3vl_8b_bf16.safetensors structural reference |
| Comfy-Org/Qwen-Image-2.1 | df94239739eef7205973a145ffa6f441f7b64e84 |
Official INT8 policy/layout and tested diffusion/VAE |
AtomicChat sources: Qwen-Image-2.1-Turbo-Abliterated-Uncensored-BF16.gguf and mmproj-Qwen-Image-2.1-Turbo-Abliterated-Uncensored-BF16.gguf. The original architecture and weights originate from Qwen/Qwen3-VL-8B-Instruct.
Credit to Qwen, AtomicChat, Comfy-Org, the ComfyUI Native Quantizer and comfy-kitchen projects for the original weights, modifications, compatible layout and conversion/runtime implementations. The silveroxides/convert_to_quant project was researched for format compatibility; its learned-rounding converter did not produce these files.
The source's Not For All Audiences classification is retained. “Abliterated” identifies the inherited language-weight modification; no image uncensoring effect or new safety/refusal behavior has been established by this release.
SHA-256 checksums
34153e12be478cac8debc0619c6c1d4f4dd53cd42c333346585e268d5ca7dde3 qwen3vl_8b_abliterated_int8_convrot.safetensors
e9e4e7f0c046ac95749628a50234faf1fd3334a62a7a34273e6c578464095cbc qwen3vl_8b_abliterated_bf16.safetensors
License and notices
The Qwen3-VL weights and AtomicChat-derived encoder are distributed under Apache-2.0. This distribution reconstructs the 73 modified language tensors in the Comfy-Org checkpoint layout and supplies an INT8 ConvRot derivative. Both include the complete vision tower. The diffusion model and VAE are separate dependencies and are not redistributed here.
ComfyUI and comfy-kitchen retain their own licenses. Conversion used the Apache-2.0 ComfyUI Native Quantizer backend; conversion software is not bundled in this model-only repository.
Apache License 2.0 — full text
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.