- Open source AI models have nearly closed the performance gap with proprietary systems like ChatGPT and Claude, with only a 3% difference, while costs have dropped up to 50x in three years.
- Open models power about one-third of real-world AI usage but capture only 4% of revenue, indicating value isn't flowing back into the open ecosystem.
- China and East Asia lead in open source AI adoption at 89%, treating it as a national strategy, while many countries are enacting new AI strategies and restricting foreign data processing.
- 79% of developers use open models, but only 51% have deployed them in production, highlighting a gap in infrastructure and tooling rather than model quality.
- The key competitive layer is now the 'agentic harness'—software controlling what AI can see, remember, and do—with few guardrails and high user consent fatigue.
- Mozilla's report emphasizes the need to invest in infrastructure, tooling, and governance to ensure open AI can scale in the public interest.