Open Source LLM Statistics and Trends (2026) (openllmstack.com)

Open source large language models (LLMs) have rapidly evolved, transforming from research interests into essential components of production systems within just three years. They now power a variety of applications including coding assistants and enterprise workflows, with their adoption outpacing proprietary models in certain sectors. A report highlights that as of 2025, there are over 2 million public model repositories on Hugging Face, with open source models capturing about one-third of token usage on platforms like OpenRouter. The DeepSeek model stands out, achieving impressive milestones including 1 billion downloads and dramatically impacting the market by significantly reducing the costs associated with LLM inference. This growth is significant for the AI/ML community as it indicates a shift toward more accessible and customizable AI solutions. The open source sector is not only catching up in terms of performance—competing closely with closed models—but is also showing a diverse geographical contributor base, especially from China. The continued decline in inference costs, driven by advances in technology and operational efficiency, suggests that open models will become increasingly feasible and appealing for developers and businesses. As such, the landscape is becoming more competitive, leading to greater innovation and democratization in AI model availability.