Open-weight models hit 62% of enterprise tokens, under 9% of spend

Open-weight models reached approximately 62% of token volume on Vercel's AI Gateway by August 22, 2026, up from 11% in April 2026, while capturing under 9% of spending. Closed models retain roughly two-thirds of dollars spent. Vercel's July data shows open-weight models at 36% of token volume, with Anthropic collecting 65.1% of all gateway spending on only 29.8% of volume, pricing its tokens at 4.4 times the average of all other labs. Chinese-built open models are a subset of the broader open-weight figure, representing about 29% of tokens in June, a number frequently conflated with the total open-weight share. On OpenRouter, top Chinese models processed 36.39 trillion tokens in the week ending July 19, versus 7.39 trillion for top US models. Z.ai's GLM 5.2 had the fastest adoption of any model Vercel tracked in 2026, with daily token volume growing 27x and customer count growing 80x in its first full week. Xiaomi's MiMo-V2.6-Pro-RL, released in October 2026, ranks first among open models on Artificial Analysis's benchmark; it reuses the same base models as V2.5 but applies significantly more reinforcement learning training compute.