Citi Report Reveals Open-Source Model Surge
New data from Citi (Citibank) shows enterprises are accelerating cuts to AI spending, directly boosting market adoption of low-cost open-source models. According to Reuters, Citi's report indicates that on the AI aggregation platform OpenRouter, the share of tokens processed by open-source models jumped from 34% in January to 65% in June this year — nearly doubling in just five months. This shift reflects a sharp increase in enterprise cost sensitivity.
The 20x Cost Gap
Citi highlights that cost disparity is the core driver behind the rise of open-source models. Data reveals that certain Chinese open-source large models, while continuously narrowing the performance gap with top U.S. models, charge as low as $0.18 per million tokens. In contrast, the average fee for current industry-leading large models is approximately $4 — a more than 20-fold difference. This massive price advantage makes open-source models the preferred choice for budget-constrained enterprises, particularly in scenarios with frequent data processing and inference tasks.
Market Implications and Outlook
The rapid increase in open-source adoption could profoundly reshape the AI industry landscape. On one hand, leading closed-source model providers face pricing pressure and may need to adjust their fee structures to retain market share. On the other hand, the growing competitiveness of Chinese open-source models may give them a stronger voice in the global AI ecosystem. Citi's report suggests that if the trend of enterprise AI spending migrating from high-end customization to low-cost standardization continues, it will accelerate AI technology adoption but may also compress profit margins for some vendors. Investors should closely monitor subsequent changes in enterprise AI budget allocation.