WqYuu/zhongshu-9b
WqYuu/zhongshu-9b is a 9 billion parameter multimodal model that integrates a DARE TIES merged language model with the vision encoder from Qwen3.5-9B. The language model component was derived from a merge of Qwen3.5-9B-LLM and HuatuoGPT-3-9B-LLM. This research model is designed for image-text tasks, leveraging its vision capabilities, though it has not yet undergone multimodal GRPO training.
Loading preview...
zhongshu-9b: A 9B Multimodal Research Model
zhongshu-9b is a 9 billion parameter multimodal model developed by WqYuu, combining a unique language model with the vision capabilities of Qwen3.5-9B. Its language model component is a DARE TIES merge of Qwen3.5-9B-LLM and HuatuoGPT-3-9B-LLM, while its vision encoder is directly sourced from Qwen3.5-9B.
Key Characteristics
- Multimodal Architecture: Integrates a powerful language model with a robust vision encoder for image-text understanding.
- Merged Language Model: The text generation capabilities are built upon a DARE TIES merge of two distinct 9B language models.
- Qwen3.5 Vision Encoder: Utilizes the unchanged vision encoder from Qwen3.5-9B, enabling strong visual processing.
- Research Focus: Currently a research model, it has not yet undergone multimodal GRPO training, indicating ongoing development.
Usage and Considerations
This model is intended for research into multimodal applications, particularly those involving image and text inputs. Users should employ AutoProcessor and AutoModelForImageTextToText from a compatible Transformers version (e.g., 5.9.0 or newer). It's important to note that this is a research tool, and any outputs, especially in sensitive domains like medical answers, require independent review and should not be used for diagnostic purposes.