Mozilla has published its inaugural State of Open Source AI report, based on a survey of 950+ developers. Key findings: the performance gap between open and proprietary AI models has narrowed to just 3%, while costs have dropped up to 50x in three years. Despite 79% of developers using open models, only 51% have deployed them in production versus 63% for closed models, pointing to missing infrastructure and tooling rather than model quality issues. China and East Asia lead global open source AI adoption at 89%. The report also highlights that the critical battleground has shifted from models to the agentic layer — the software controlling what AI can see, remember, and do — which currently operates with few guardrails. Open models power roughly a third of real-world AI usage but capture only 4% of revenue, raising sustainability concerns for the open ecosystem.
