Jev was used by about 35 percent of paying teams within its first two weeks on Vercel’s AI Gateway, according to the company’s production index, nearly double Sonnet 5.5’s comparable reach. The decision model is designed to choose among preset outputs. The newly published figures describe usage within Vercel’s routing service, which separates team adoption, token volume and spending rather than treating them as one measure.
According to TechCrunch, Vercel, a developer-platform company that operates the AI Gateway model-routing service, has published production usage figures for Jev. About 35 percent of paid teams used the model in its first two weeks, nearly twice the comparable adoption for Sonnet 5.5. Vercel calls it the fastest adoption on its gateway. The count covers traffic through that service.[1]
Jev selects from a fixed set of possible responses, a task designed to run quickly instead of yielding a long answer. Vercel describes moving the classification of potentially harmful commands from GPT-6 Luna to Jev in its own application. The production index separates the number of teams using a model from token volume and spending, recording distinct measures of activity within the gateway.[1]
More than 70 percent of gateway tokens are generated by open models, whose share of spending is about 25 percent. Over the comparison period, expenditure on leading closed-model laboratories rose by roughly ten times. The gateway data exclude direct requests to model companies and deployments routed elsewhere. Vercel is both the service operator and a user of models within this channel.[1]
