Open-Weight AI Usage Rises While Closed Models Capture More Spending

Summary

The article compares open-weight and closed-source AI models using activity on OpenRouter, an API platform that connects users to inference providers. It argues that growing use of open models does not necessarily mean frontier labs are losing enterprise customers or revenue: organizations may prefer to pay providers to manage the technical work of running models themselves.

The evidence cited is a divergence between usage and estimated spending. Since mid-May, open-weight models’ share of tokens on the platform rose to 75%, while closed-source models continued to account for much more estimated spend. This suggests that token volume alone may be a poor proxy for revenue or commercial strength. The figures reflect one platform, and the article provides no methodology for estimating spend or comparison with the wider market. Its conclusions are therefore limited and do not establish the relative profitability or long-term competitiveness of either model category.

Key ideas

• Open-weight models gained token share on the platform while closed-source models retained higher estimated spending.
• Token usage and revenue can move differently when providers charge different prices.
• Some enterprises may pay closed-model providers to avoid operating inference infrastructure themselves.
• The cited platform data may not represent the entire AI market.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.