Mozilla publishes its first State of Open Source AI report

Sources: State of Open Source AI report, Mozilla blog, HN discussion
Summary: Mozilla published its inaugural State of Open Source AI report on 2026-07-14, built on a survey of more than 950 developers and its own analysis. It reports the performance gap between open and top proprietary models has narrowed to about 3 percent, with inference costs down up to 50x over three years. It states open models now serve roughly a third of real-world usage but capture about 4 percent of revenue, that 79 percent of surveyed developers use open models while only 51 percent run them in production versus 63 percent for closed models, and that closed models still lead on reasoning, long-context retrieval, and complex agentic tasks. It puts China and East Asia ahead on open-model adoption at about 89 percent.
Comments: HN commenters criticized the report page readability and animation, and debated whether commoditized open weights erode the economics of frontier labs that carry the training cost.
Why it matters: The survey frames the remaining barriers to open-model adoption as integration, maintenance, and deployment rather than raw capability, which is the gap teams hit when moving open models to production.