Vercel data: TypeSafe AI's Jev is its fastest-adopted model ever

Data published by Vercel on Thursday, Oct. 15, offers the first hard look at how much of the Jev hype turned into real usage: within two weeks of its limited release, the decision model built by TypeSafe AI was used by 35% of paid teams on Vercel's AI gateway, nearly double the reach of the second most popular model, Sonnet 5.5. TechCrunch reported the figures on Oct. 16.

Because Vercel's AI gateway is a unified API service similar to OpenRouter, the numbers only capture traffic routed through that gateway. TechCrunch notes the view is necessarily limited and is especially likely to undercount major labs such as OpenAI and Anthropic, which offer their own APIs.

Jev was published in limited release on Sept. 15 as a lightweight tool for choosing between a few predetermined outputs rather than generating open-ended text. Limiting the outputs lets it run cheaper and faster than general models like Sonnet, which makes it well suited to classifying large quantities of data quickly.

The approach drew significant buzz in the weeks after launch, and inspired competing offerings from OpenAI and Amazon, but until now there was little hard data to back up the good vibes. Vercel was an early adopter itself, using Jev to replace GPT-6 Luna for classifying harmful commands.

The same dataset shows open source models still gaining ground: they now account for more than 70% of the tokens processed through Vercel's gateway and nearly 25% of total spend.

Growing open source share does not mean frontier models are shrinking, though. Spending on frontier labs grew tenfold even as it accounted for a smaller share of overall spend, which TechCrunch reads as evidence that frontier models and open alternatives may turn out to be complementary rather than purely competitive.

What to watch next is whether Jev-style narrow-output models reproduce the same adoption curve on other inference and distribution platforms, and how gateway-based measurements shape the way the industry ranks models. After all, whoever controls the distribution channel holds data closest to the real shape of the market.