The Jev AI model from TypeSafe AI has become the fastest-adopted model in Vercel AI Gateway’s history. It reached 35.3% of paid teams within its first two weeks, nearly doubling Claude Sonnet 5.5’s 18.1% reach. The figures come from Vercel’s October AI Gateway Production Index, published Thursday.
Released in limited access on September 15, Jev makes rapid decisions between predefined choices instead of generating open-ended text. This approach makes it suitable for data classification, software automation and AI agent monitoring. Its specialised design can reduce processing costs and response times compared with larger language models.
Jev AI Model Sets a New Adoption Record
Jev received more requests than any other model on Vercel’s gateway during September, despite launching midway through the month. Of teams that tried Jev in its first week, 55.6% returned the following week. That ranked fifth for early user retention among the launches Vercel measured.
However, high adoption did not translate into substantial spending. Jev handled about 1% of September’s gateway tokens but accounted for just 0.15% of spending. Its smaller, cheaper requests explain why widespread usage generates relatively little revenue compared with larger models.
Vercel also adopted Jev early, using it for command-safety classification. As TechAmerica.ai previously reported, the model replaced a larger AI system in that workflow. Its early momentum has encouraged competing approaches, including OpenAI’s Decisions API and Amazon’s Strands Decider 2B.
Open-Weight AI Models Capture More Gateway Traffic
Vercel’s report also shows growing demand for open-weight models. They processed 71.5% of gateway tokens during September and accounted for approximately 23% of total spending. Their spending share has more than quadrupled since June.
Meanwhile, spending on frontier models increased tenfold since December, even as their share of total spending declined. Open-weight model spending grew more than seventyfold over the same period. Overall gateway spending rose approximately thirteenfold, while the average cost per token fell 47%.
The figures suggest developers increasingly use different AI models for different workloads. Vercel found that 68% of teams increasing their open-weight spending also increased spending on closed-weight alternatives. This indicates that lower-cost systems can complement frontier models rather than replace them.
Vercel’s data covers traffic routed through its AI Gateway, not the entire AI market. Direct API usage through providers such as OpenAI and Anthropic falls outside that measurement. Nevertheless, Jev’s rapid adoption shows developer demand for specialised decision models, even if their long-term commercial impact remains uncertain.
