Mozilla: Open AI Models Narrow Gap to Frontier Models
The performance gap between open AI models and closed frontier models has narrowed to four months, according to a Mozilla report. Many companies are shifting to cheaper open models.
A new report from Mozilla indicates a significant reduction in the performance gap between open-source AI models and proprietary, closed-source frontier models.
The report highlights that a leading open model, Moonshot AI's Kimi K3, trails Anthropic's Fable 5 closed model by only three points in performance while costing just 30 percent of the latter.
Raffi Krikorian, Mozilla's Chief Technology Officer, stated that closed models offer premium value in specific areas like expert professional work and intensive retrieval tasks.
The findings suggest that organizations should consider open models as the default for most workloads due to their cost-effectiveness, reserving more expensive closed models for highly specialized tasks.