Part of the Deep Dive: Data Governance Webinar Series
China’s open source ecosystem is entering a period of rapid growth, driven by the rise of open-source AI models, infrastructure projects, and an expanding developer base. In this keynote, Kaiyuanshe Co-founder Emily Chen and Chairwoman Nadia will share data-backed insights into developer trends, open source contributions, and the technologies at the core of this transformation.
Video transcript
Speakers: Emily Chen and Nadia Jiang (Board Chair of Kaiyuanshe)
Introduction to Kaiyuanshe
Emily Chen: We’re excited to share data-driven observations on trends and insights from China’s open source ecosystem in the AI era. Our goal is to build a bridge between open source communities in the East and West.
I’m Emily Chen, an open source contributor since 2004. I co-founded Kaiyuanshe in 2014 and have been actively involved in open source communities.
Nadia: I’m Nadia, board chair of Kaiyuanshe. I work at Ant Group and am an active open source contributor and Apache Software Foundation member. I’ve helped many global open source foundations build their local communities in China.
Kaiyuanshe, founded in 2014, is an open source alliance in China known as “the home of open source people.” Our vision is rooted in China, contributing globally, and advancing open source as a way of life. As a volunteer-driven community, we operate on three principles: contribution, consensus, and collegiality.
Our mission focuses on: open source governance, international bridge-building, community development, and project incubation. We collaborate with communities, universities, enterprises, developers, and governments, and have been OSI’s first China-based member since 2016.
Global Open Source Landscape
Emily: Following the Open Source Congress in Brussels this summer, we’ve seen tremendous global attention on China’s open source growth—from the DeepSeek moment to the rapid wave of open source models and AI infrastructure projects.
Kaiyuanshe publishes an annual China Open Source Report capturing key trends. Today we’re sharing findings from our data-driven research.
Key Statistics
- Global developers reached a new high by July 2025: over 150 million worldwide
- 22.8 million open source developers actively contribute on GitHub
- China has over 12 million developers total, with about 4 million active open source developers
- China ranks second globally in open source contributions
Top 10 Global Open Source Technologies
- AI large language models
- Cloud infrastructure
- Front-end and interaction technologies
- Programming languages and development tools
Open Source LLM Development Landscape
Nadia: In September, my team at Ant Group, collaborating with Kaiyuanshe, released “Open Source LLM Development 2025: Landscape, Trends and Insights.”
Major LLM Trends (January-August 2025)
Open Weight Dominance: Most Chinese LLMs are open weight, demonstrating that Chinese enterprises are fully embracing open source.
Scaling Through MoE Architecture: Flagship models like DeepSeek, Qwen, and Kimi adopted Mixture of Experts (MoE) architecture with sparse activation, enabling trillion-parameter models like GPT-4, Claude Opus, and O3.
Reinforcement Learning: Models like DeepSeek R1 combine large-scale pre-training with RL-based post-training, significantly improving performance in automatic reasoning, complex decision-making, and knowledge inference. Strong reasoning has become a defining feature of flagship LLM releases in 2025.
Multimodal Models: Multimodal capabilities have gone mainstream across the ecosystem.
Open Source LLM Ecosystem Analysis
We identified 114 prominent open source projects across 22 technical domains using the OpenRank evaluation system, which calculates project influence based on community collaboration.
Key Findings:
- 62% of these projects launched after the GPT moment (October 2022)
- 12 projects started in 2025 alone
- Average of nearly 30,000 GitHub stars per project
- Over 360,000 GitHub accounts have contributed through issues or pull requests
Top Keywords: AI (126 mentions), LLM (98), Agent (81), followed by data, learning, search, model, OpenAI, framework, Python, and MCP.
Developer Contributions by Country
Among 124,351 developers with location data:
- US: 37.4% of contributions
- China: 18.7% of contributions
- Germany: 6.5% (third place)
US and Chinese developers are the core driving forces behind the open source LLM development ecosystem, together accounting for over 55% of total contributions.
Contributions by Technical Domain
AI Infrastructure & UI: Chinese developers account for over 60% of contributions; Germany contributes less than 4%
AI Data: More globally distributed, with Poland, Norway, France, and the Netherlands all ranking in the top 10
AI Agents: US developers contribute 24.6%, Chinese developers 21.5%—showing relatively higher Chinese engagement in this space
Conclusion
Nadia: Today’s presentation represents a small slice of our full report. For the complete landscape and detailed data, visit the URLs provided. Follow us on Twitter and LinkedIn for the latest trends and insights on China’s open source and AI ecosystem.
Emily: Kaiyuanshe is here to help bridge the West and East in the open source community. Contact us to learn more about China’s open source ecosystem.
Sources: Ant Open Source and X-Lab OpenDigger, UN Open Source Innovation Report, Kaiyuanshe Annual China Open Source Report