Scarf launches daily open source AI adoption leaderboard

Scarf has introduced a public leaderboard that tracks real-world adoption of AI model providers and inference platforms by measuring organization-level open source package downloads. The index aims to give developers, investors and researchers a directional view of which AI technologies are spreading fastest across the open source ecosystem.

Why it matters: - The leaderboard offers a new signal for AI adoption beyond benchmarks, model launches and product announcements. - Scarf is measuring whether organizations are actually pulling AI-related open source packages, which can better reflect developer and enterprise interest. - The rankings may help developers decide which providers and runtimes to support, while giving analysts a way to watch shifts in open source AI usage over time.

What happened: - Scarf launched the Open Source AI Popularity Leaderboard, or OSS AI Leaderboard, on August 11, 2026. - The index updates daily and ranks AI model providers and inference platforms by the number of distinct organizations Scarf sees downloading related open source packages. - Scarf processes billions of open source package downloads each day and built the leaderboard from its catalog of nearly 700,000 tracked packages. - Avi Press, Scarf founder and CEO, said open source adoption provides another way to understand which AI technologies developers and organizations are integrating deeply.

The details: - The leaderboard counts each organization once per provider per day, no matter how many related packages it downloads or how often it downloads them. - Scarf says that method reduces the effect of large deployments and automated systems on the rankings. - The company publishes two views: model providers and inference platforms. - Model providers are companies or organizations that develop and publish AI models accessed through open source packages, including open-weight and proprietary models. - Inference platforms include hosted services, software platforms and local runtimes that execute AI models or provide access to them. - Companies can appear in both categories if they offer both models and infrastructure. - Scarf combines data from Scarf Gateway with information from partner package registries to create a cross-registry view. - The first snapshot shows OpenAI with a significant lead in package adoption. - Amazon holds two of the top four inference platform spots through Amazon Bedrock and Amazon SageMaker. - Ollama ranks sixth among inference platforms and sits ahead of several major cloud offerings, including Azure. - Scarf says that result points to strong demand for local model execution and open source AI infrastructure. - Kimi ranks lower in package adoption despite producing prominent open-weight models. - The ranking highlights a gap between technical achievement and downstream adoption.

Between the lines: - The leaderboard is designed to measure distribution and pull in the ecosystem, not raw model usage. - The index does not track API requests, token volume, provider revenue, model performance or which models users run through general-purpose inference platforms. - An Ollama package download counts toward Ollama, but the data does not reveal which model an organization used afterward. - The framing suggests Scarf sees package downloads as an early indicator of broader AI infrastructure adoption.

What's next: - Scarf plans to expand coverage as it identifies more provider-associated packages and gains access to additional data sources. - Organizations seeking granular data, historical datasets or a live feed can contact help@scarf.sh. - The OSS AI Leaderboard is available now.