jed351/Gemma3-4B-ChatVector_SFT-from-IT_and_IT

VISIONConcurrency Cost:1Model Size:4.3BQuant:BF16Ctx Length:32kPublished:Jan 23, 2026Architecture:Transformer Cold

jed351/Gemma3-4B-ChatVector_SFT-from-IT_and_IT is a 4.3 billion parameter large language model developed by hon9kon9ize, based on Google's Gemma3-4B. This model has undergone continuous pre-training (CPT), supervised fine-tuning (SFT), and model merging to enhance its capabilities, particularly for Cantonese language understanding and generation. It excels at interpreting nuanced Cantonese phrases and cultural contexts, offering more accurate responses compared to its base model. Its primary use case is providing precise information and explanations in Cantonese.

Loading preview...

CantoneseLLM v2 Beta

CantoneseLLM v2 Beta, developed by hon9kon9ize, is a 4.3 billion parameter language model built upon Google's Gemma3-4B. This model has been significantly enhanced through a combination of continuous pre-training (CPT), supervised fine-tuning (SFT), and strategic model merging with the original Gemma3 4B. The goal is to establish it as the best-in-class open-weight model for Cantonese.

Key Capabilities

  • Superior Cantonese Understanding: Demonstrates improved accuracy and contextual awareness when interpreting complex Cantonese phrases and cultural nuances, as shown in direct comparisons with the base Gemma3 4B-it model.
  • Enhanced Response Quality: Provides more precise and culturally appropriate explanations for colloquial Cantonese terms.
  • Optimized for Cantonese Dialogue: Fine-tuned to act as an AI assistant specifically designed for accurate and concise information delivery in Cantonese.

Good for

  • Cantonese Language Applications: Ideal for chatbots, virtual assistants, or educational tools requiring deep understanding and generation of Cantonese.
  • Cultural Context Interpretation: Useful for scenarios where explaining Cantonese idioms and their cultural significance to non-native speakers is crucial.
  • Research and Development: Serves as a strong foundation for further advancements in Cantonese-specific LLMs, with plans for additional data integration and long-context capability enhancements.

Finetunes
1 models