Measuring Open Model Inference Share
Researcher Nathan Lambert published an introductory research note examining the challenges and methodologies involved in measuring the inference share of open-source AI models. Understanding inference share helps industry analysts and developers track the real-world adoption and economic impact of open models compared to proprietary alternatives. The note is targeted at paid subscribers and focuses on the nuances of quantifying open-source model usage within the broader AI economy.
## BACKGROUND
AI inference refers to the phase where a trained model processes inputs to generate predictions or responses, which typically drives the ongoing operational costs of AI applications. Open-source or open models allow users to download, modify, and self-host the software, contrasting with closed API-based services.