The Pulse
Vercel Says Jev Reached 35.3% of Paid Teams in Two Weeks
Vercel’s October AI Gateway index says TypeSafe AI’s Jev reached 35.3% of paid teams during its first 14 days on the service, the highest launch reach in the index’s history. Jev accounted for 0.27% of gateway spending over that period, des
AI.info Team · Written with AI ·
35.3% reach, 0.27% of spending
Jev reached 35.3% of paid teams using Vercel AI Gateway during its first 14 days, according to Vercel’s October 15 production index. The TypeSafe AI model ranked first for launch reach among the September releases tracked in the report, nearly doubling the 18.1% recorded for Claude Sonnet 5.5. Yet Jev accounted for only 0.27% of Gateway spending during its launch window.
The gap points to two different kinds of adoption: many teams tried Jev, but their use represented a small share of the money routed through Vercel’s service. Vercel also says Jev received more requests than any other model in September, despite launching halfway through the month. The figures cover Vercel AI Gateway traffic, not all use of AI models across companies or services.
“Whether a model launch is successful depends on what you measure: how broadly it’s tried, how much usage and spend it receives, or what share of its initial users return.”
Amelia Charles and Harpreet Arora, Vercel
More teams tried Jev than returned
Vercel’s scorecard shows that 55.6% of teams that tried Jev in its first week used it again in the second. That placed Jev fifth for return rate among the September launches with at least two weeks of data. OpenAI’s GPT-6 Luna led that measure at 64.8%, while Sonnet 5.5 posted 60.5%.
The comparison separates initial reach from repeat use. Jev drew the broadest first-two-week audience in Vercel’s table, but its return rate did not lead the field. Vercel’s published figures do not establish why teams tried Jev or why some did not return in week two.
A model built for structured decisions
Vercel introduced Jev as a probabilistic decision model for structured decisions in software. Rather than producing open-ended prose, it evaluates questions supplied with an application’s context and returns typed choices, scores or true-or-false answers with probabilities. That output can feed directly into code.
Vercel lists possible uses including selecting an agent’s next tool, deciding whether a workflow should retry or stop, scoring risk, and sending uncertain cases for human review. The company’s launch post also relays TypeSafe AI’s claim that, in its own workflow evaluations, Jev was up to 194 times faster and 445 times cheaper than language models. Those are vendor-reported evaluation results, not measurements from Vercel’s adoption index.
Vercel’s figures cover its own gateway
The October index is based on anonymized, aggregate traffic routed through AI Gateway through September 2026. Vercel defines launch reach using external paid teams active from the start of the launch month through the data cutoff, and counts each model’s first 14 days. Spending is estimated at published list prices, so the report says actual bills may differ.
That scope makes the result a measure of Jev’s uptake among Vercel Gateway customers, not a market-wide adoption rate. The index records a striking contrast—35.3% reach and 0.27% of launch-window spend—but does not show whether Jev’s early trial translated into sustained use beyond the first return window.