We analyzed the AI model ecosystem on Hugging Face, examining over +23,000 models and +6,000 datasets from 100 organizations. Our data shows a relationship between community engagement and practical adoption: while platforms such as Hugging Face maintain extensive follower networks and high engagement metrics, individual models show dramatically different deployment patterns based on their origin. Commercial models from organizations such as Meta, Microsoft and Nvidia seem to outperform academic alternatives in download volume, suggesting that practical AI implementation relies heavily on commercially developed solutions, despite research institutions contributing numerous available models.
Our visualisation also shows a significant correlation between institutional backing and deployment metrics, spanning over +300 million model downloads and +447,000 model likes, as well as +9.5 million dataset downloads and +44,000 dataset likes. Furthermore, few models successfully achieve both high community engagement and substantial download numbers, indicating a pronounced gap between innovation and implementation. Individual developers and organizations must bridge this divide by identifying and adapting research innovations for practical applications. Otherwise, there might be a widening disparity between cutting-edge academic research and deployed AI solutions, reinforcing existing divisions within the ecosystem.
However, some organizations demonstrate distinct approaches to bridging this divide. Ai2 (Allen Institute for AI) exemplifies the research-to-practice model with +700 models, representing sustained academic output aimed at practical applications, while Meta's approach with +2,000 models demonstrates how commercial organizations can achieve both scale and deployment success by combining research capabilities with market-focused development.
Lastly, Hugging Face occupies a distinct position within its own landscape, functioning primarily as a hosting platform rather than developing its own models. The platform generates high follower counts and community engagement while its individual models receive fewer downloads compared to commercial model developers like Meta or Microsoft.
Note: While download and engagement metrics provide valuable insights into usage patterns, these numbers represent tracked interactions through APIs and may not capture all user activity.
Source: Hugging Face