Mozilla’s inaugural State of Open Source AI report reveals that open models have nearly closed the performance gap with proprietary systems while significantly reducing operational costs for global developers.

Key Points

• Open source AI models now perform within 3% of top proprietary systems like ChatGPT and Claude.
• Costs for running open models have decreased by up to 50 times over the past three years.
• While 79% of developers use open models, only 51% have successfully deployed them into production environments.
• China and East Asia lead global adoption at 89%, treating open source AI as a core national strategy.
• Open models currently power one-third of real-world AI applications but generate only 4% of total industry revenue.
• The report identifies "agentic harnesses"—the software layer controlling AI actions—as the most critical area for future development and governance.

Why it Matters

The shift toward open source AI offers organizations the ability to own their infrastructure rather than renting proprietary services from major tech providers. However, a lack of mature deployment tooling and sustainable revenue models threatens to limit the long-term viability of open ecosystems compared to closed, scalable alternatives.