New data from Vercel's AI Gateway shows open-weight models have overtaken proprietary rivals in both usage and spending. By September, they handled 71.5% of all tokens through the gateway, with the majority share first reached in August. Their share of spending more than quadrupled over three months, rising from one in twenty gateway dollars in June to nearly one in four in September.
Proprietary US models still capture the largest single chunks of spending: Claude Opus 5.5 took 11.03% of September spend and GPT-6 Astra took 4.86%. But Chinese open-weight models are climbing quickly. DeepSeek V4.1 Flash ranked third in spending at 4.19%, while Moonshot AI's Kimi and Z.ai's GLM-5.3 also captured meaningful shares of tokens and spend.
The source notes a key wrinkle: more than two-thirds of September spending on Chinese models went to US providers serving those models, up from under a tenth in December 2025. That makes proposed US restrictions on open-weight models politically complicated, and newly proposed Chinese export controls on AI-related technology could add further uncertainty for US businesses that host or rely on Chinese models. Gateway data does not capture direct vendor access or on-premises hosting, so the full market may differ from these trends.