Scarf has launched the Open Source AI Popularity Leaderboard, a public index updated daily that ranks AI model providers and inference platforms based on package downloads by organisations. The index is based on Scarf’s processing of billions of open-source package installations. From nearly 700,000 monitored packages, Scarf identifies AI-related packages and maps them to their respective model and inference providers. Each organisation is counted once per day, regardless of how many packages it downloads, helping prevent large-scale installations or automated activity from distorting the rankings. The leaderboard separates model providers, which create and distribute AI models using open-source packages, from inference platforms, which provide services for hosting or running those models. Organisations can appear in either or both categories depending on their products. Scarf collects the underlying data through its gateway and package-registry integrations. The first leaderboard snapshot shows OpenAI leading package adoption, while Amazon occupies two of the top four positions among inference platforms through Amazon Bedrock and Amazon SageMaker. Ollama ranks sixth among inference platforms, while other providers show differences between their technological capabilities and package-level adoption. The leaderboard measures package adoption rather than direct model usage, and does not account for API requests, token counts, provider revenue, model performance or the specific models used for general-purpose inference. Scarf describes the index as a directional measure of open-source package adoption rather than a measure of the entire AI market.