A decision-only model that never writes a sentence reached roughly 13% of Vercel's paid AI Gateway teams in its first 24 hours.
The fastest-adopted AI model on Vercel’s AI Gateway does not write essays, poems or code. It answers questions with a yes, a no, a pick from a list, or a score.
That model is Jev, built by TypeSafe AI. According to Vercel’s gateway data, it reached about 13% of the platform’s paid teams within 24 hours of launch. That figure outpaced every major chatbot-style model the gateway tracks.
The data comes from Vercel’s AI Gateway, a unified API service. Developers plug into one interface and switch between providers without rewriting their apps.
Jev launched on September 15, 2026. Within its first day, its adoption among paid gateway teams was double that of the GPT-5.6 family at the same point. It was also more than six times the uptake of Fable 5.1 over the same window.
Vercel later formalized the result. Its AI Gateway Index, published on October 15, 2026, confirmed Jev as the fastest-adopted model on the gateway. The index also named it the most requested offering for September.
Jev is described as a probabilistic decision model. Instead of generating free-form text, it returns structured, typed answers. Those come in three flavors: Boolean (true or false), Choice (pick one option from a set), and Score (a numeric rating).
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Since Jev produces no text, the gateway charges no output-token fees for it. Input costs run approximately $0.04 to $0.042 per million tokens.
TypeSafe says Jev can deliver results up to 194 times faster and 445 times cheaper than traditional language models. Those are TypeSafe’s own figures, and they describe best-case comparisons rather than every workload.
Vercel’s engineers offered a more grounded data point from their own systems. They swapped an existing safety classifier for Jev and reported speed gains of between 5 and 18 times. They also said accuracy improved.
TypeSafe did not launch quietly. On the same day Jev went live, the company announced a $40 million seed round led by DCVC.
The model is available through Vercel’s AI Gateway and is hosted by TypeSafe. Within three days of launch, it had been integrated into Cloudflare and LangChain.
Early use cases point toward operational work. Risk assessment and workflow decisions are the kinds of production tasks where Jev is being applied.
A large share of real-world AI work is classification: is this message spam, does this transaction look risky, which queue should this ticket go to. Many teams have been using full-size language models for those jobs because they were the tool at hand.
There are caveats worth keeping in view. The adoption figures come from one gateway, so they reflect Vercel’s customer base rather than the whole industry. The headline speed and cost multiples are vendor claims. Day-one trials also differ from sustained production use.