base_model: AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-BF16
license: apache-2.0
tags:

• gptq
• 4-bit
• quantized
• gptqmodel
• foem
• qwen3
• uncensored
  pipeline_tag: text-generation

Qwen3.6-27B-AEON-Ultimate-Uncensored — GPTQ-Pro FOEM 4-bit g128

Overview

Qwen3.6-27B-AEON-Ultimate-Uncensored-GPTQ-Pro-FOEM-4bit-g128 is a GPTQ-quantized checkpoint intended for efficient GPU inference, published by groxaxo.
It is intended for open-source evaluation, reproducible experimentation, and compatible local or
hosted inference workflows. The wording below is deliberately limited to what can be verified
from this repository's metadata and artifacts.

The repository name identifies a behavior-modified or reduced-filtering lineage. That label describes the source or conversion history; it is not a guarantee of unrestricted behavior in every prompt or runtime. Test outputs carefully before sharing or deploying them.

At a glance

What is included

• *.safetensors (5 files)
• config.json
• generation_config.json
• tokenizer.json
• tokenizer_config.json
• processor_config.json
• chat_template.jinja
• quantize_config.json
• Additional configuration, tokenizer, processor, or shard files (16 visible artifacts total)

Quick start

vLLM (documented configuration)

vllm serve groxaxo/Qwen3.6-27B-AEON-Ultimate-Uncensored-GPTQ-Pro-FOEM-4bit-g128 \
  --quantization gptq_marlin \
  --dtype float16 \
  --trust-remote-code

This command is taken from the repository documentation. Adjust tensor parallelism, context
length, and cache settings to match your hardware and vLLM version.

Compatibility and responsible use

• Use a runtime that explicitly supports this format, architecture, and modality.
• Keep configuration, tokenizer, processor, projection, and weight files from the same revision together.
• Review the source model card and license before redistribution or deployment.
• Hardware needs depend on parameter count, context length, cache precision, quantization, and concurrency.
• Report reproducible issues with the runtime version, hardware, launch command, and a minimal example.

Quantization or conversion changes numerical behavior, memory use, and throughput relative to the source checkpoint; validate quality on your own workload.

Generated outputs may be inaccurate or unsuitable for a given use case. Users are responsible for
testing behavior, applying appropriate safeguards, and complying with applicable licenses and laws.

Quantized version of AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-BF16
using GPTQModel with FOEM (First-Order Error Minimization) enhancement.

Quantization Recipe

Quantized: language_model.layers linear modules only
(attn projections, MLP gate/up/down)

Preserved in full BF16:

• model.visual.* — entire vision tower (333 tensors)
• lm_head.weight
• model.language_model.embed_tokens.weight
• All norm layers, RoPE, and multimodal glue

Perplexity Comparison (WikiText-2)

Evaluated on identical settings (512 ctx / 256 stride, wikitext-2-raw-v1 test set):

Only 1.6% perplexity degradation at 4-bit — an excellent result for W4G128 GPTQ. FOEM + activation-weighted MSE preserved language model fidelity across all 64 transformer layers.

Usage

from gptqmodel import GPTQModel, BACKEND

model = GPTQModel.load(
    "AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-GPTQ-Pro-FOEM-4bit-g128",
    device="cuda:0",
    backend=BACKEND.AUTO,
)

Quantization Details

• Tool: GPTQModel v6.1.0-dev
• Calibration: 64 samples from WikiText-2
• Hardware: 3× RTX 3090/3060 (CUDA_VISIBLE_DEVICES=0,1,2)
• Duration: ~112 minutes
• gc_mode: on_stage_end (VRAM-safe for large VLMs)
• Offload: disk offload enabled during quantization

About the Base Model

This quantization is based on AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-BF16, an uncensored variant of Qwen3 27B with vision capabilities (architecture: Qwen3_5ForConditionalGeneration).