p-e-w/gemma-3-12b-it-heretic-v2

Model Overview

p-e-w/gemma-3-12b-it-heretic-v2 is a decensored version of Google's Gemma 3 12B instruction-tuned model, developed using the Heretic v1.1.0 tool. This model maintains the core capabilities of the Gemma 3 family, which are lightweight, state-of-the-art open models built from the same research as Gemini models.

Key Differentiators

• Decensored Performance: Compared to the original google/gemma-3-12b-it, this Heretic v2 variant shows a significant reduction in refusals, dropping from 97/100 to 7/100, while maintaining a low KL divergence of 0.0995.
• Multimodal Capabilities: The model processes both text and image inputs (normalized to 896x896 resolution, encoded to 256 tokens each) and generates text outputs.
• Extended Context Window: Features a large 128K token input context window, supporting complex and lengthy interactions.
• Multilingual Support: Offers robust support for over 140 languages, enhancing its applicability across diverse linguistic contexts.

Intended Use Cases

This model is well-suited for a range of applications, particularly where reduced content moderation is desired:

• Content Creation: Generating creative text formats, marketing copy, and email drafts.
• Conversational AI: Powering chatbots and virtual assistants.
• Text Summarization: Creating concise summaries of documents and research papers.
• Image Understanding: Extracting, interpreting, and summarizing visual data for text communications.
• Research and Education: Serving as a foundation for VLM and NLP research, language learning tools, and knowledge exploration.