SAVRN
Search Contact SAVRN

Open-weight model

Qwen-Image-2.1-Text-Encoder-Heretic

by Pottokao pottokao/Qwen-Image-2.1-Text-Encoder-Heretic

Qwen-Image-2.1-Text-Encoder-Heretic is an open-weight model from Pottokao, released under Apache License 2.0. It has 8.8B parameters and a 262,144-token context. At 16-bit it needs about 21 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index. It draws 799 downloads a month.

The text encoder of Qwen/Qwen-Image-2.1 (a Qwen3-VL-8B-Instruct) with refusal behaviour removed via Heretic directional ablation. Drop-in replacement for the stock text encoder. Weights are bf16, same shapes, same parameter count — nothing else was changed.

Parameters8.8B
Context262,144
Weights35.1 GB
Licenseapache-2.0
AccessOpen weights
Monthly Downloads799

Runs On

What it takes to serve Qwen-Image-2.1-Text-Encoder-Heretic (8.8B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.

PrecisionWeightsMemory neededCheapest setupPer hourAlso fits
16-bit 17.5 GB 21.0 GB 1x MI300X (192 GB)
Vultr
$1.85 1x H100 $1.99 · 1x MI325X $2.00
8-bit 8.8 GB 10.5 GB 1x MI300X (192 GB)
Vultr
$1.85 1x H100 $1.99 · 1x MI325X $2.00
4-bit 4.4 GB 5.3 GB 1x MI300X (192 GB)
Vultr
$1.85 1x H100 $1.99 · 1x MI325X $2.00

Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Sep 23, 2026.

Qwen-Image-2.1-Text-Encoder-Heretic on every accelerator the SAVRN Index prices, at every precision

Model Card

By Pottokao, published under apache-2.0, revision 106be9554084.

Qwen-Image-2.1 Text Encoder — Heretic (Abliterated)

Not affiliated with, or endorsed by, Alibaba / Qwen. Community derivative (refusal-ablated) of Qwen/Qwen3-VL-8B-Instruct — the model Qwen-Image-2.1 uses, unmodified, as its text encoder. Qwen releases that model under Apache-2.0, so this derivative is redistributed under Apache-2.0 (see LICENSE and NOTICE).

The text encoder of Qwen/Qwen-Image-2.1 (a Qwen3-VL-8B-Instruct) with refusal behaviour removed via Heretic directional ablation.

Drop-in replacement for the stock text encoder. Weights are bf16, same shapes, same parameter count — nothing else was changed.

Results

Refusals KL divergence
Original (measured baseline) 100/100 0 (by definition)
This model 5/100 0.0220

Measured by Heretic on mlabonne/harmful_behaviors (test split) for refusals and mlabonne/harmless_alpaca for KL divergence — i.e. lower refusals and lower distribution shift on benign inputs.

Independent verification

Refusal rate and general capability were re-checked with a separate script (different code, different refusal keyword set) rather than trusting the optimiser's own numbers:

Read the full model card (801 words)

Configuration

Architecture
Qwen3VLForConditionalGeneration
Context length (tokens)
262,144
Layers
36
Hidden size
4,096
Feed-forward size
12,288
Attention heads
32
Key/value heads
8
Head dimension
128
Vocabulary size
151,936
Model type
qwen3_vl

Identity and Version

Repository
pottokao/Qwen-Image-2.1-Text-Encoder-Heretic
Publisher
Pottokao
Task
Not stated by the source
Modality
Other
Library
Not stated by the source
Parameters
8.8B parameters
Languages
Not stated by the source
Revision
106be95540846f5905485637472144091b83e524
First published
2026-09-20
Last updated
2026-09-23

Files and Weights

16 files, 35.1 GB in total. The weights are 5 files totalling 35.1 GB in safetensors.

Weights5 files · 35.1 GB
Configuration4 files · 71.0 KB
Tokenizer2 files · 11.4 MB
Documentation3 files · 19.0 KB
Other1 file · 5.3 KB
Repository1 file · 1.6 KB
Every file
FileTypeSizeSHA-256
model-00001-of-00004.safetensorsWeights4.9 GB 6e217f5e222d
model-00002-of-00004.safetensorsWeights4.9 GB 0226ae289505
model-00003-of-00004.safetensorsWeights5.0 GB 9c62f604a704
model-00004-of-00004.safetensorsWeights2.7 GB 8516dd606abf
qwen3vl_8b_bf16_heretic.safetensorsWeights17.5 GB b1f17ffe6e04
config.jsonConfiguration1.6 KB
generation_config.jsonConfiguration213 B
model.safetensors.index.jsonConfiguration67.8 KB
processor_config.jsonConfiguration1.4 KB
LICENSEDocumentation11.4 KB
NOTICEDocumentation812 B
README.mdDocumentation6.8 KB
chat_template.jinjaOther5.3 KB
.gitattributesRepository1.6 KB
tokenizer.jsonTokenizer11.4 MB be75606093db
tokenizer_config.jsonTokenizer418 B

License and Download

License
apache-2.0
Access
Open weights, no gate
Download size
35.1 GB
Download from Pottokao

Released by Pottokao through its official repository on Hugging Face. Read the license.

Built From

Memory Requirements

PrecisionWeights in memory
As published35.1 GB
16-bit17.5 GB
8-bit8.8 GB
4-bit4.4 GB

Weights only, from the published parameter count; the key-value cache and runtime add to this.

Built on This Model

Questions About Qwen-Image-2.1-Text-Encoder-Heretic

How much GPU memory does Qwen-Image-2.1-Text-Encoder-Heretic need?

About 21 GB at 16-bit and 5.3 GB at 4-bit: the weights (8.8B parameters) plus a working margin. A long context needs more.

What is the cheapest GPU to run Qwen-Image-2.1-Text-Encoder-Heretic on?

At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.

Can I use Qwen-Image-2.1-Text-Encoder-Heretic commercially?

Yes. Qwen-Image-2.1-Text-Encoder-Heretic is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.

What is Qwen-Image-2.1-Text-Encoder-Heretic's context length?

262,144 tokens, from the maximum position embeddings in its published configuration.