Toward open-source foundation model ecosystem: Impact evaluation framework and promotion mechanism

Open-source foundation models cultivate complex innovation ecosystems that render traditional, project-centric evaluation frameworks inadequate. To address this gap, our study develops and validates a three-level framework for assessing ecosystem-level impact and identifying its enhancement mechanisms. Grounded in technology diffusion theory, we conduct a mixed-methods analysis of 14 leading models, using data from Hugging Face, GitHub, and X (formerly Twitter). Our findings reveal that while the initial impact of these models is balanced, significant gaps emerge at the secondary (derivative innovation) and tertiary (global influence) levels. We term this challenge the “climbing effect”—the difficulty of transitioning impact across these levels—and identify specific technical and strategic control points that facilitate this progression. Theoretically, this study shifts the unit of analysis from individual projects to broader ecosystems, challenges the assumption of “smooth diffusion,” and introduces control point theory to the open-source context. Practically, our findings offer actionable strategies for developers and an evidence-based framework for policymakers to foster a more prosperous open-source AI landscape.