Model Overview

This model, Jackrong/GPT-5-Distill-Qwen3-4B-Instruct, is a 4 billion parameter instruction-tuned conversational LLM built upon the Qwen/Qwen3-4B-Instruct-2507 base model. Its unique characteristic lies in its training methodology: it undergoes Supervised Fine-Tuning (SFT) on ShareGPT data and incorporates knowledge distillation from LMSYS GPT-5 responses. This process aims to imbue the model with the conversational style and quality of GPT-5, offering high-quality, natural-sounding dialogues while maintaining low computational overhead.

Key Capabilities

• Lightweight and Efficient: Approximately 4 billion parameters, ensuring fast inference and reduced resource consumption.
• GPT-5 Distillation-Style Responses: Designed to mimic the conversational fluency and helpfulness observed in GPT-5.
• Highly Conversational: Optimized for engaging chatbot-style interactions and rich dialogue flows.
• Multilingual Support: Seamlessly handles both Chinese and English inputs and outputs.
• Extended Context Window: Supports a maximum context length of 32,768 tokens.

Recommended Use Cases

• Casual chat in Chinese and English.
• General knowledge explanations and reasoning guidance.
• Code suggestions and basic debugging assistance.
• Writing assistance, including editing, summarizing, and rewriting tasks.
• Role-playing conversations with appropriately designed prompts.