TypeSafe's Jev hits 35% of Vercel teams at 0.27% of spend

TypeSafe's Jev reached 35% of Vercel's paid teams in its first two weeks on the AI Gateway, the fastest adoption on record for a model there, while consuming 0.27% of gateway spend over the same window. TechCrunch reported the launch, drawing on Vercel's own October production index.

The gap between those two numbers is the whole story. Jev is a decision model, built to return typed answers like true/false, pick-one, or score-it, rather than free-form prose, so each call burns a fraction of the tokens a general-purpose chat model would. In Vercel's own words, "Jev was the fastest-adopted model in AI Gateway history, reaching 35% of paid teams in its first two weeks, but it accounted for just 0.27% of gateway spend over that same period." The report ranks Jev's 55.6% retention as the fifth-highest among new models.

Vercel used the launch to replace GPT-6 Luna, which it had been running to classify harmful commands, with Jev. Luna itself, OpenAI's frontier model, "reached nearly three out of every twenty paid teams in its first two weeks, in third place behind Jev and Sonnet 5.5," and captured under 1% of spend despite the month's highest retention at 64.8%.

The wider shift in the index is the open-weight tilt. "Open-weight models now process more than seven in ten gateway tokens and account for nearly one in four dollars," the report says, with that share of spend having more than quadrupled since June. September's total gateway spend was "about 13 times December's, an increase of 1,180%, while the average price per token fell 47%." Our open-source tracker has logged 358 stories in that lane over the last 90 days.