This year has seen rapid advancements in China's open-source large model landscape, with an array of significant iterations. Commencing with DeepSeek/R1 in January, the industry has witnessed the release of Alibaba's Qwen series, Moonshot, Zhipu AI's Z.ai, and MiniMax, each unveiling new models that offer open-source or open-weight versions for free download and modification. These initiatives have significantly bolstered the development of global open-source models. The swift progress of China's open-source large models has put pressure on US vendors traditionally committed to closed-source strategies. Consequently, OpenAI responded by introducing its first open-source model, gpt-oss, in August. This move underscores the global impact of open-source models and signals the emergence of a new competitive paradigm in the AI sector.
