prithivMLmods / ultragemma4-e2b-heretic-uncensored

huggingface.co
Total runs: 2.8K
24-hour runs: 0
7-day runs: 2.5K
30-day runs: 2.5K
Model's Last Updated: June 26 2026
image-text-to-text

Introduction of ultragemma4-e2b-heretic-uncensored

Model Details of ultragemma4-e2b-heretic-uncensored

ultragemma4-e2b-heretic-uncensored

Reasoning-capable language model modified using the Heretic abliteration toolkit

Abliteration E2B Parameters Reasoning Uncensored

ultragemma4-e2b-heretic-uncensored is a reasoning-capable language model built on top of google/gemma-4-E2B-it and modified using the heretic abliteration toolkit. The model applies refusal-direction analysis and targeted weight-space interventions to reduce internal refusal behaviors while preserving instruction-following, reasoning capabilities, and general conversational performance.

Important

This model is intended strictly for research and learning purposes. Due to reduced internal refusal mechanisms, it may generate sensitive or unrestricted content. Users assume full responsibility for how the model is used. The authors and hosting platform disclaim any liability for generated outputs.

Note

This model is experimental and may generate unexpected behaviors or artifacts in certain scenarios.

Key Highlights
  • Heretic-Based Abliteration : Modified using the Heretic toolkit to identify and alter refusal-related representations within the model.
  • Reduced Refusal Behavior : Optimized to minimize internal refusal tendencies while maintaining instruction-following capabilities.
  • Gemma 4 Backbone : Built directly on top of google/gemma-4-E2B-it .
  • Reasoning-Oriented Performance : Preserves multi-step reasoning and analytical capabilities after abliteration.
  • Research-Focused Release : Designed for alignment research, model behavior analysis, and evaluation of refusal-direction modifications.
  • Efficient E2B Deployment : Suitable for local inference, research environments, and optimized deployment setups.
Model Files
File Name Quant Type File Size File Link
ultragemma4-e2b-heretic-uncensored.BF16.gguf BF16 9.27 GB Download
ultragemma4-e2b-heretic-uncensored.F16.gguf F16 9.27 GB Download
ultragemma4-e2b-heretic-uncensored.Q2_K.gguf Q2_K 2.98 GB Download
ultragemma4-e2b-heretic-uncensored.Q3_K_L.gguf Q3_K_L 3.27 GB Download
ultragemma4-e2b-heretic-uncensored.Q3_K_M.gguf Q3_K_M 3.19 GB Download
ultragemma4-e2b-heretic-uncensored.Q3_K_S.gguf Q3_K_S 3.1 GB Download
ultragemma4-e2b-heretic-uncensored.Q4_0.gguf Q4_0 3.35 GB Download
ultragemma4-e2b-heretic-uncensored.Q4_K_M.gguf Q4_K_M 3.42 GB Download
ultragemma4-e2b-heretic-uncensored.Q4_K_S.gguf Q4_K_S 3.35 GB Download
ultragemma4-e2b-heretic-uncensored.Q5_0.gguf Q5_0 3.58 GB Download
ultragemma4-e2b-heretic-uncensored.Q5_K_M.gguf Q5_K_M 3.62 GB Download
ultragemma4-e2b-heretic-uncensored.Q5_K_S.gguf Q5_K_S 3.58 GB Download
ultragemma4-e2b-heretic-uncensored.Q6_K.gguf Q6_K 3.83 GB Download
ultragemma4-e2b-heretic-uncensored.Q8_0.gguf Q8_0 4.93 GB Download
ultragemma4-e2b-heretic-uncensored.mmproj-bf16.gguf mmproj-bf16 987 MB Download
ultragemma4-e2b-heretic-uncensored.mmproj-f16.gguf mmproj-f16 987 MB Download
ultragemma4-e2b-heretic-uncensored.mmproj-q8_0.gguf mmproj-q8_0 557 MB Download
Intended Use
  • Alignment Research : Studying refusal-direction analysis and behavior modification techniques.
  • Model Evaluation : Benchmarking reasoning, instruction-following, and safety-related behaviors.
  • Red Teaming : Analyzing model responses under reduced-refusal conditions.
  • Local Deployment : Running compact Gemma 4 models in research and experimentation environments.
  • Abliteration Studies : Exploring the effects of targeted weight-space modifications on model behavior.
Limitations & Risks

Important Note : This model intentionally reduces built-in refusal mechanisms.

  • Sensitive Content Risk : May generate unrestricted, controversial, or unsafe outputs.
  • User Responsibility : Requires careful and ethical use.
  • Experimental Modifications : Behavior may differ significantly from the original model.
  • Alignment Trade-offs : Reduced refusal behavior may impact safety filtering and response constraints.
  • Potential Artifacts : Certain prompts may expose unexpected outputs resulting from the abliteration process.
Acknowledgements
  • google/gemma-4-E2B-it : Gemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling text and image input (with audio supported on small models) and generating text output. This release includes open-weights models in both pre-trained and instruction-tuned variants. Gemma 4 features a context window of up to 256K tokens and maintains multilingual support in over 140 languages.

    Featuring both Dense and Mixture-of-Experts (MoE) architectures, Gemma 4 is well-suited for tasks like text generation, coding, and reasoning. The models are available in four distinct sizes: E2B, E4B, 26B A4B, and 31B. Their diverse sizes make them deployable in environments ranging from high-end phones to laptops and servers, democratizing access to state-of-the-art AI.

  • Heretic : Fully automatic censorship removal framework for language models. This project was used to perform the refusal-direction analysis and ablation procedures that form the foundation of this model.

Abliteration Parameters
Parameter Value
direction_index 26.34
attn.o_proj.max_weight 1.00
attn.o_proj.max_weight_position 29.29
attn.o_proj.min_weight 0.59
attn.o_proj.min_weight_distance 13.90
mlp.down_proj.max_weight 1.49
mlp.down_proj.max_weight_position 23.32
mlp.down_proj.min_weight 0.66
mlp.down_proj.min_weight_distance 7.76
Refusal Evaluation
Metric This model Original model (google/gemma-4-E2B-it)
Refusals 12/100 99/100
llama.cpp

LLM inference in C/C++ — https://github.com/ggml-org/llama.cpp

license

Gemma 4 [Apache License 2.0] — https://ai.google.dev/gemma/apache_2

Runs of prithivMLmods ultragemma4-e2b-heretic-uncensored on huggingface.co

2.8K
Total runs
0
24-hour runs
323
3-day runs
2.5K
7-day runs
2.5K
30-day runs

More Information About ultragemma4-e2b-heretic-uncensored huggingface.co Model

More ultragemma4-e2b-heretic-uncensored license Visit here:

https://choosealicense.com/licenses/apache-2.0

ultragemma4-e2b-heretic-uncensored huggingface.co

ultragemma4-e2b-heretic-uncensored huggingface.co is an AI model on huggingface.co that provides ultragemma4-e2b-heretic-uncensored's model effect (), which can be used instantly with this prithivMLmods ultragemma4-e2b-heretic-uncensored model. huggingface.co supports a free trial of the ultragemma4-e2b-heretic-uncensored model, and also provides paid use of the ultragemma4-e2b-heretic-uncensored. Support call ultragemma4-e2b-heretic-uncensored model through api, including Node.js, Python, http.

ultragemma4-e2b-heretic-uncensored huggingface.co Url

https://huggingface.co/prithivMLmods/ultragemma4-e2b-heretic-uncensored

prithivMLmods ultragemma4-e2b-heretic-uncensored online free

ultragemma4-e2b-heretic-uncensored huggingface.co is an online trial and call api platform, which integrates ultragemma4-e2b-heretic-uncensored's modeling effects, including api services, and provides a free online trial of ultragemma4-e2b-heretic-uncensored, you can try ultragemma4-e2b-heretic-uncensored online for free by clicking the link below.

prithivMLmods ultragemma4-e2b-heretic-uncensored online free url in huggingface.co:

https://huggingface.co/prithivMLmods/ultragemma4-e2b-heretic-uncensored

ultragemma4-e2b-heretic-uncensored install

ultragemma4-e2b-heretic-uncensored is an open source model from GitHub that offers a free installation service, and any user can find ultragemma4-e2b-heretic-uncensored on GitHub to install. At the same time, huggingface.co provides the effect of ultragemma4-e2b-heretic-uncensored install, users can directly use ultragemma4-e2b-heretic-uncensored installed effect in huggingface.co for debugging and trial. It also supports api for free installation.

ultragemma4-e2b-heretic-uncensored install url in huggingface.co:

https://huggingface.co/prithivMLmods/ultragemma4-e2b-heretic-uncensored

Url of ultragemma4-e2b-heretic-uncensored

ultragemma4-e2b-heretic-uncensored huggingface.co Url

Provider of ultragemma4-e2b-heretic-uncensored huggingface.co

prithivMLmods
ORGANIZATIONS

Other API from prithivMLmods