The Future of the Global Open-Source AI Ecosystem: From DeepSeek to AI+
6 hours ago
- #DeepSeek
- #Open-Source AI
- #China AI Ecosystem
- The article concludes a three-part series on China's open-source AI advancements since the 'DeepSeek Moment' in January 2025.
- Open-source is the dominant approach for Chinese AI organizations, fostering large-scale deployment and integration.
- China's organic open-source AI ecosystem includes established players like Alibaba, Tencent, ByteDance, and Baidu, as well as startups like Moonshot and Z.ai.
- Alibaba's Qwen became the most reused model on Hugging Face, with over 113k derivatives, surpassing Meta's Llama.
- Tencent shifted from borrowing to building, releasing its own models like Tencent HY, focusing on vision, video, and 3D.
- ByteDance selectively open-sourced high-value components while maintaining a competitive edge in AI applications like Doubao, which hit 100M DAU.
- Baidu re-entered the open-source ecosystem with Ernie 4.5 and renewed investment in PaddlePaddle and its AI chip Kunlunxin.
- Startups like Moonshot, Z.ai, and MiniMax quickly adapted, contributing models like Kimi K2 and GLM-4.5, with some announcing IPOs.
- Application-first companies like Xiaohongshu and Bilibili began training in-house models, leveraging domain data and low-cost open-source reasoning.
- Research institutions like BAAI and Shanghai AI Lab focused on toolchains, evaluation systems, and deployment infrastructure.
- China's AI ecosystem now features a full chain from models to deployment, supported by national compute and energy infrastructure like the 'East Data, West Compute' strategy.
- China's 2025 'AI+' action plan emphasized large-scale deployment and integration, moving beyond AGI to practical, scalable AI systems.
- DeepSeek R1 catalyzed China's AI acceleration, embedding AI deeper into industrial processes and fostering autonomous, controllable systems.
- The shift to open-source as a default assumption in system design marks a fundamental change in how models are used, making them reusable components in larger systems.
- China's AI trajectory, shaped by resource constraints, focuses on open-source, engineering efficiency, and scalable delivery, creating a self-sustaining ecosystem with global implications.