Positioning Vercel's access path for Jev runs through the AI Gateway, which hosts the model under typesafe-ai/jev, while the AI SDK and Vercel's own agent framework provide the upper-level calling and evaluation surfaces on top of it.
What it does The AI Gateway serves Jev under the model ID typesafe-ai/jev, with Zero Data Retention available for teams that need it. AI SDK 7 exposes the model through its experimental evaluate API, in which a Boolean evaluation maps onto Jev's Noul question type. Vercel's agent framework eve goes further: it sets Jev as the default evaluator for five of its features, replacing prompted LLM judgments with decision calls. A changelog entry documents the model's availability and the Zero Data Retention option. The changelog entry lists the model ID and the retention option.
Characteristics By Vercel's own account in its launch blog — a vendor-reported figure rather than an independently measured one — Jev is the fastest-adopted model in AI Gateway's history, with 13% of paid teams using it within 24 hours of launch, a figure the company says indicates how quickly decision-model workloads moved onto the platform. For developers already building on the AI SDK, Jev is a native option rather than an extra dependency.
When to use Suits teams building on Vercel or inside the AI SDK ecosystem that want decision-model calls as a native option, and agent projects on the eve framework, which get Jev-backed evaluation on five features by default. The adoption figure is vendor-reported, and the evaluate API is experimental, so watch for changes when relying on it in production setups.