license: gemma
base_model: HauhauCS/Gemma-4-E2B-Uncensored-HauhauCS-Aggressive
base_model_relation: quantized
pipeline_tag: image-text-to-text
inference: false
language:

• en
• multilingual
  tags:
• llamafile
• gguf
• gemma4
• multimodal
• vision
• audio
• uncensored
• abliterated
• quantized
• q6_k_p
• apple-silicon
• metal

Gemma-4-E2B Uncensored HauhauCS Aggressive Q6_K_P — llamafile

An unofficial, reproducible llamafile package of the Q6_K_P GGUF from HauhauCS/Gemma-4-E2B-Uncensored-HauhauCS-Aggressive.

HauHauCS claims:

No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended — just without the refusals.

These are meant to be the best lossless uncensored models out there.

Aggressive Variant

Stronger uncensoring — model is fully unlocked and won't refuse prompts. May occasionally append short disclaimers (baked into base model training, not refusals) but full content is always generated.

These are HauhauCS's claims. This package has not independently validated the claimed refusal behavior, capability preservation, or uncensoring results.

The single executable contains:

• Gemma-4-E2B-Uncensored-HauhauCS-Aggressive-Q6_K_P.gguf;
• mmproj-Gemma-4-E2B-Uncensored-HauhauCS-Aggressive-f16.gguf for supported multimodal input;
• llamafile 0.10.5; and
• terminal chat, browser chat, and a localhost API server.

The model and f16 projector were embedded without retraining, merging, re-quantization, or compression.

Intended use

This package is for people who want one downloadable executable containing a small, local Gemma 4 E2B model, its multimodal projector, a terminal chat interface, a browser chat interface, and a localhost API server. It is convenient for offline or self-contained macOS/Linux use where installing a separate inference runtime is undesirable.

Because everything is in one file, you can keep it on a USB drive and bring a self-contained local AI with you. According to HauhauCS's claim above, this aggressive variant is fully unlocked and does not refuse prompts. The destination computer still needs a supported 64-bit platform and enough RAM, and the drive or filesystem must permit executable files.

Included artifact

Embedded components and reproducibility tools

Run

On macOS or Linux:

chmod +x Gemma-4-E2B-Uncensored-HauhauCS-Aggressive-Q6_K_P.llamafile
./Gemma-4-E2B-Uncensored-HauhauCS-Aggressive-Q6_K_P.llamafile

The terminal chat starts directly. While it is running, open the browser chat at:

http://127.0.0.1:8080/

The OpenAI-compatible API is available under:

http://127.0.0.1:8080/v1

The server is bound to localhost by default. Do not expose it publicly without authentication, TLS, and appropriate network controls.

Packaged defaults

• model: /zip/Gemma-4-E2B-Uncensored-HauhauCS-Aggressive-Q6_K_P.gguf;
• projector: /zip/mmproj-Gemma-4-E2B-Uncensored-HauhauCS-Aggressive-f16.gguf;
• context: 8,192 tokens;
• sampler: temperature 0.6, top-p 0.95, top-k 20, min-p 0.0;
• Jinja chat templates enabled;
• supported multimodal input enabled through the embedded f16 projector; and
• additional command-line arguments pass through after the embedded defaults.

To override the context size, pass --ctx-size when launching; for example:

./Gemma-4-E2B-Uncensored-HauhauCS-Aggressive-Q6_K_P.llamafile --ctx-size 32768

Larger contexts increase memory use. The upstream GGUF advertises a native context length of 131,072 tokens, but practical limits depend on the runtime, KV-cache settings, and available memory.

Multimodality and architecture

The f16 projector is paired with the main GGUF inside the executable. The HauhauCS card describes Gemma 4 E2B as natively multimodal for text, image, video, and audio and states that vision/audio use requires the projector. Actual media handling still depends on the input client and the capabilities exposed by llamafile 0.10.5.

The original lineage is google/gemma-4-e2b-it. Its E2B architecture has 35 layers, 512-token sliding-window attention interleaved with full attention, and a 128K-class context window (131,072 tokens in this GGUF's metadata). Google's original card describes 2.3B effective parameters, or 5.1B including embeddings.

Provenance

• modified/uncensored model and GGUF source: HauhauCS/Gemma-4-E2B-Uncensored-HauhauCS-Aggressive at repository revision da8593c3e407afcd3e7da94ff2d69d77e2a28a48
• original model lineage: google/gemma-4-e2b-it
• quantization: Q6_K_P, with an importance matrix according to the HauhauCS card
• projector precision: f16
• package runtime: Mozilla llamafile 0.10.5
• license metadata: gemma, following the direct source repository; see LICENSE

Notes

• This is an unofficial repackaging, not a new model release.
• The embedded projector makes the package multimodal; image, audio, and video handling still depends on client/UI and embedded-runtime support.
• The upstream repository describes the model as aggressively uncensored. This package does not independently reproduce or validate its refusal or capability claims.
• The packaged 8,192-token default is conservative. Override it with --ctx-size N, such as --ctx-size 32768, if your memory budget permits.
• This full-runtime artifact is larger than 4 GiB, so it is intended primarily for macOS and Linux rather than direct execution on Windows.
• Hugging Face hosted inference is disabled because this repository distributes a standalone executable rather than a standard hosted-inference checkpoint.