Economies of Open Intelligence: Tracing Power & Participation in the Model Ecosystem
Shayne Longpre, Christopher Akiki, Campbell Lund, Atharva Kulkarni, Emily Chen, Irene Solaiman, Avijit Ghosh, Yacine Jernite, Lucie-Aimée Kaffee
2025-12-04
Summary
This paper analyzes how the world of openly available AI models has changed over time, specifically looking at who is creating them, how powerful they are becoming, and how openly the information about them is shared.
What's the problem?
The AI world is rapidly evolving, and it's becoming unclear who controls the development of these models. There's a concern that a few big companies might dominate, or that models are becoming less transparent about the data they were trained on. It was difficult to get a clear picture of these trends because there wasn't a comprehensive dataset tracking model downloads and characteristics.
What's the solution?
Researchers gathered a huge dataset – over 850,000 models and 2.2 billion downloads – from the Hugging Face Model Hub, spanning from June 2020 to August 2025. They then analyzed this data to identify shifts in which companies and groups are creating popular models, how the models themselves are changing in terms of size and capabilities, and whether they are truly open source or more restricted. They also created a public dashboard to allow others to track these changes.
Why it matters?
This research shows a shift in power away from major US tech companies like Google and Meta towards independent developers, communities, and now, Chinese companies. It also highlights that models are getting much larger and more complex, with new techniques being used, but that transparency about the data used to build them is decreasing. Understanding these trends is crucial for ensuring a diverse and open AI ecosystem, and for monitoring potential risks associated with concentrated power or lack of transparency.
Abstract
Since 2019, the Hugging Face Model Hub has been the primary global platform for sharing open weight AI models. By releasing a dataset of the complete history of weekly model downloads (June 2020-August 2025) alongside model metadata, we provide the most rigorous examination to-date of concentration dynamics and evolving characteristics in the open model economy. Our analysis spans 851,000 models, over 200 aggregated attributes per model, and 2.2B downloads. We document a fundamental rebalancing of economic power: US open-weight industry dominance by Google, Meta, and OpenAI has declined sharply in favor of unaffiliated developers, community organizations, and, as of 2025, Chinese industry, with DeepSeek and Qwen models potentially heralding a new consolidation of market power. We identify statistically significant shifts in model properties, a 17X increase in average model size, rapid growth in multimodal generation (3.4X), quantization (5X), and mixture-of-experts architectures (7X), alongside concerning declines in data transparency, with open weights models surpassing truly open source models for the first time in 2025. We expose a new layer of developer intermediaries that has emerged, focused on quantizing and adapting base models for both efficiency and artistic expression. To enable continued research and oversight, we release the complete dataset with an interactive dashboard for real-time monitoring of concentration dynamics and evolving properties in the open model economy.