Skip to content

EthenEthenEthen

LoginPlatformTry Ethen

Open Source Model Profile · Ryex

Floppa-12B-Gemma3-Uncensored

Floppa-12B-Gemma3-Uncensored is a 12.19B-parameter Gemma 3 multimodal fine-tune from Ryex for uncensored Japanese-English translation and image description.

Publisher
Ryex
Task
image-text-to-text
Model type
gemma3
License
gemma
Library
vllm
Publication status
Accepted · not indexed

Model overview

Floppa-12B-Gemma3-Uncensored is published by Ryex as a Gemma3 image-text-to-text fine-tune. The captured configuration identifies Gemma3ForConditionalGeneration and Safetensors metadata reports 12,187,325,040 parameters. According to the model card, it is based on Gemma 3 12B Instruct for uncensored translation and image description, and card data records gemma.

Recorded capabilities

Gemma 3 12B multimodal base

According to the model card, the model is a fine-tuned multimodal model on Gemma 3 12B Instruct with a SigLIP vision encoder.

Uncensored translation and description

The publisher says it performs Japanese-English translation and image description with zero sanitization, including slang, profanity, and cultural nuance.

QLoRA Floppa Mix

According to the model card, 4-bit QLoRA with Unsloth was merged to bfloat16 after training on about 10.5k rows with toxic-text and vision-anchor mixes.

vLLM-optimized use

According to the model card, the model is optimized for vLLM with a documented image-plus-prompt generation example.

Use cases in the source record

  • Japanese-English translation experiments that need the publisher's documented unsanitized handling of slang, profanity, and nuance.
  • Image-description trials with vLLM using the documented image-plus-prompt input format.

Limitations and unknowns

  • No evaluation results were extracted from this record.
  • No context-window value was extracted from this record.
  • Uncensored and non-refusal behavior is a publisher-stated design choice, not an Ethen-verified safety finding.
  • Provider state is historical snapshot data and should be refreshed before being presented as current.

Source and provenance

Source: Ryex/Floppa-12B-Gemma3-Uncensored

Captured: Unknown. Processed: 2026-09-07T19:35:51.536096+00:00.

Floppa-12B-Gemma3 (Uncensored) Floppa-12B is a fine-tuned multimodal model based on Google's Gemma 3 12B Instruct . It has been specialized for high-fidelity uncensored translation (Japanese <-> English) and unrestricted image description . Unlike the base model, Floppa does not refuse to describe explicit, violent, or "spicy" imagery, and it translates slang, profanity, and cultural nuances with zero sanitization. Model Description Base Model: google/gemma-3-12b-it Architecture: 12B Dense Transformer with SigLIP Vision Encoder (Hybrid Attention). Training Method: 4-bit QLoRA with Unsloth, merged to bfloat16. Objective: Behavior mod…

F001F002F003F004F005F006F007F009F010F011F012F013F014F015F016F017F018