A recent report by Mozilla highlights that the performance gap between the most advanced artificial intelligence (AI) models developed by U.S. tech companies and the best open-source models from Chinese firms has narrowed significantly. According to the "State of Open Source AI" report, published on September 15, the difference is now just 4.4 months in terms of development progress. This suggests that open-source AI models are rapidly catching up with their proprietary counterparts in terms of capabilities. The report emphasizes that open-source AI models should be the default choice for most organizations, given their cost-effectiveness and growing performance. One example is Moonshot AI’s Kimi K3, an open-source model that performs nearly as well as Anthropic’s Fable 5, a closed-source model from the U.S. Kimi K3 scores just three points lower on the Artificial Analysis Intelligence Index, a benchmark that evaluates AI capabilities, while costing only 30 percent as much as Fable 5. Despite this progress, closed-source models still hold an edge in specific areas. According to Raffi Krikorian, Mozilla’s chief technology officer, closed models perform better in tasks such as expert-level professional work, high-intensity data retrieval, and handling long text passages. These advantages make closed models more attractive for certain specialized tasks, even though open models are becoming increasingly competitive. Krikorian noted that the decision to use a closed model is often based on the specific workload rather than a broad organizational preference. This means that while open-source models may be sufficient for many tasks, some businesses may still opt for closed models when dealing with complex or high-stakes projects. The report underscores the growing importance of open-source AI in the broader landscape of artificial intelligence development.