Open weight takes 71.5% of tokens and 23% of spend

AI Gateway’s October 2026 Production Index reports rapid growth and diversification in enterprise AI usage, with September gateway spending rising roughly 13-fold from December even as the average token price fell 47%. Open-weight models processed more than 70% of tokens and nearly one-quarter of spending, with their spending share increasing more than fourfold since June, although frontier-model spending also grew tenfold. Teams increasingly switch providers and models, with nearly half changing their most-used model and almost one-third changing their primary lab during September. Chinese-origin models generated substantial revenue through US-based providers, which handled nearly 70% of spending on those models. September launches showed contrasting measures of success: TypeSafe AI’s Jev achieved record adoption, reaching 35% of paid teams in two weeks but producing little spend because of its low-cost, low-token decision-model design; Anthropic’s Opus 5.5 captured 11% of gateway spending in its first two weeks; and OpenAI’s GPT-6 Luna led launch retention while GPT-6 Astra attracted a smaller audience but more spending. The report is based on anonymized aggregate Vercel AI Gateway traffic and estimates spending using published list prices.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.