license: apache-2.0
tags:

• text-generation-inference
• moe
• uncensored
• abliterated
• unfiltered
• unredacted
• refusal-ablated
• vllm
• pytorch
• bf16
• max
• alignment-modified
• reasoning
• agent
  base_model:
• google/gemma-4-26B-A4B-it
  model-index:
• name: gemma-4-31B-it-Uncensored-MAX
  results:
  • task:
    type: image-text-to-text
    metrics:
    • type: abliteration_rate
      value: 94.4
      name: Abliteration Rate

language:

• en
  pipeline_tag: image-text-to-text
  library_name: transformers

gemma-4-26B-A4B-it-Uncensored-MAX

gemma-4-26B-A4B-it-Uncensored-MAX is an optimized release built on top of huihui-ai/Huihui-gemma-4-26B-A4B-it-abliterated. This version focuses on updated shard sizing, repository optimization, and compatibility improvements for the latest Transformers releases, while preserving the MoE architecture and reasoning capabilities of the original Gemma model. The result is a powerful 26B parameter Mixture-of-Experts language model designed for efficient inference, stable deployment, and modern ecosystem integration.

[!IMPORTANT]
This model is intended for research and learning purposes only. Any content generated by this model is used at the user's own risk. The authors and hosting page disclaim any liability for outputs produced by this model. Users are responsible for ensuring safe, ethical, and lawful usage.

Evaluation Report (Self-Reported)

Note: The evaluation was conducted using 500 selected harmful test prompts to measure model refusal behavior. These results are self-reported and may vary depending on benchmark setup and evaluation methodology.

Key Highlights

• Latest Transformers Compatibility
  Re-sharded and optimized for improved compatibility with recent Transformers releases.

• Optimized Model Sharding
  Updated shard structure for better storage handling, download reliability, and inference efficiency.

• Stable Inference Pipeline
  Improved packaging for consistent loading and generation behavior across environments.

• 26B MoE Architecture (A4B)
  Built on gemma-4-26B-A4B-it, leveraging Mixture-of-Experts design for scalable reasoning capacity.

• Improved Deployment Stability
  Designed for smoother inference across different hardware configurations and runtimes.

• Preserved Model Behavior
  No modifications to weights or architecture; behavior remains consistent with the base model lineage.

Base Model Signatures:

This model has been re-sharded and optimized for the latest Transformers version from the base model:
https://huggingface.co/huihui-ai/Huihui-gemma-4-26B-A4B-it-abliterated

Quick Start with Transformers

pip install -U transformers

from transformers import Gemma4ForConditionalGeneration, AutoProcessor
import torch

model = Gemma4ForConditionalGeneration.from_pretrained(
    "prithivMLmods/gemma-4-26B-A4B-it-Uncensored-MAX",
    torch_dtype="auto",
    device_map="auto"
)

processor = AutoProcessor.from_pretrained(
    "prithivMLmods/gemma-4-26B-A4B-it-Uncensored-MAX"
)

messages = [
    {
        "role": "user",
        "content": [
            {"type": "text", "text": "Explain how transformer models work in simple terms."}
        ],
    }
]

text = processor.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)

inputs = processor(
    text=[text],
    padding=True,
    return_tensors="pt"
).to("cuda")

generated_ids = model.generate(**inputs, max_new_tokens=256)

generated_ids_trimmed = [
    out_ids[len(in_ids):] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
]

output_text = processor.batch_decode(
    generated_ids_trimmed,
    skip_special_tokens=True,
    clean_up_tokenization_spaces=False
)

print(output_text)

Intended Use

• Multimodal and Language Research
  Studying large-scale transformer and MoE behavior.

• Red-Teaming & Evaluation
  Testing robustness across complex and adversarial prompts.

• High-Performance Deployment
  Running large models on optimized GPU or distributed inference setups.

• Research Prototyping
  Experimentation with scalable transformer architectures.

Limitations & Risks

Important Note: This model inherits the behavior and limitations of its base model.

• Output Variability
  Responses may vary depending on sampling configuration and prompt structure.

• Resource Requirements
  A 26B MoE model requires significant GPU memory and optimized inference strategies such as quantization or tensor parallelism.

• Deployment Constraints
  Performance depends heavily on hardware configuration and runtime optimization.

• General Model Limitations
  May produce incorrect, incomplete, or inconsistent outputs in complex scenarios.