All stories
Jev Momentum Source-backed

Jev tops Vercel's model adoption chart while taking just 0.27% of spend

Vercel says Jev reached 35.3% of paid teams in its first 14 days, nearly twice Sonnet 5.5's reach. Broad trial and a large inference bill are different measures.

By AITrending2 min read
Sources (2)

Why now Vercel published its October Production Index on October 15, and TechCrunch reported the Jev adoption result on October 16. The new event is a usage report about September launches, not a new Jev release.
Vercel's chart of paid-team reach during models' first 14 days, with Jev above 35 percent and historical comparison launches below it.
Vercel's historical adoption chart for AI Gateway paid teams. It includes earlier launches beyond the September scorecard and measures gateway reach, not worldwide market share or independently audited usage. Resized without cropping. Source: Vercel

A decision model wins on reach

Jev was the fastest-adopted model in Vercel AI Gateway history, according to Vercel’s October Production Index. The company reports that TypeSafe AI’s decision model reached 35.3% of paid teams in its first 14 days, compared with 18.1% for Claude Sonnet 5.5, second in the report’s September launch comparison. The accompanying historical chart includes earlier launches too.

The report, published October 15 and covered by TechCrunch the next day, lists Jev’s gateway launch as September 17. These are retrospective adoption figures for an existing model, separate from Jev’s recent 1.14 feature update.

Jev returns constrained decisions rather than open-ended prose. Vercel says it received the most requests of any model in September, despite joining the gateway halfway through the month. Of teams that tried it in week one, 55.6% returned in week two, the fifth-highest return rate among the compared launches.

Lots of teams, a small share of the bill

The reach result did not translate into a similar spending share. Vercel reports that Jev accounted for 0.27% of gateway spending during its first two weeks. Its full-September spend share was 0.15%; those figures cover different windows and should not be interchanged. Vercel attributes the small bill to requests that are cheaper, faster and use fewer tokens than typical language-model calls.

This is a useful signal of interest in specialized decision models, with a defined boundary: it describes external paid teams using Vercel’s gateway. It does not measure worldwide enterprise adoption, usage through labs’ direct APIs or decision accuracy.

Vercel calculates spending from published list prices, so actual customer bills may differ. Its reach denominator includes paid teams active from the launch month’s start through the data cutoff. A team trying Jev is not evidence that it replaced its other models; the report measures adoption, spending and return rates separately.

Related reading