China Unicom has released the industry's first open source general thinking chain model for central enterprises - the Yuanjing Think Chain Model, and has opened sourced on multiple platforms such as GitHub, Moddai, and Shizhi. This model demonstrates strong slow thinking ability and multidisciplinary and multi-scenario general reasoning ability, surpassing models such as OpenAI GPT-4 and Deepseek V3 in mainstream list evaluation. Its task adaptive and difficulty adaptive strategies realize the adaptive slow thinking of the model, and optimize the matching of answer length with the difficulty of the question through reinforcement learning, improving accuracy and practicality. The Yuanjing Thinking Chain Model has been successfully applied to many fields such as physics, chemistry college entrance examination questions, 24-point games, logical reasoning, etc., showing huge application prospects.
Today, the Yuanjing Thinking Chain Model released by China Unicom marks a major breakthrough in the field of artificial intelligence and has also injected new vitality into the development of artificial intelligence technology in China and even the world. The open source of this model also reflects China Unicom's determination to actively promote the innovation and application of artificial intelligence technology and contribute to future economic and social development. Visit link: GitHub: https://github.com/UnicomAI/Unichat-32B-c1.git; Magic: https://www.modelscope.cn/UnicomAI/Unichat-qwen2.5-32B-c1.git; Shizhi: https://wisemodel.cn/models/UnicomLLM/Unichat-qwen2.5-32B-c1