Vercel says Jev was AI Gateway's top model in September

Vercel posted that Jev became AI Gateway's fastest-adopted model ever, and September's most-requested model. That is the entire claim. No request counts, no growth rate, no second place, no window beyond the month.

Vercel

@vercel

X

Jev became AI Gateway’s fastest-adopted model ever, and September’s most-requested model.

Diogo Almeida reposted it with one word, "Jevtember".

What the claim is and is not

This is a share-of-traffic statement about one gateway, made by that gateway. It says something about what developers routing through Vercel's AI Gateway chose to call in September. It does not say how many calls that was, how far ahead of the next model Jev finished, or what those calls were doing.

It is also not a quality result. Nothing in the post measures accuracy, latency, cost per million tokens or calibration. "Most-requested" rewards the model people reach for first, which is driven by defaults, docs, pricing pages and whatever shipped most recently, not by a win on anyone's eval. If you are deciding whether to put Jev in a production path, this post gives you popularity and nothing else.

For people building on it

Adoption signals like this one do tell you something real, which is that a lot of other teams are hitting the same endpoint you are considering, so the integration path is well worn and the failure modes are likely to be public. That is worth a little. It is not worth as much as a number from your own workload.

It lands alongside a separate adoption data point from a different source, the Ramp spend data that earlier showed Jev picking up a point of market adoption over a month. Two adoption readings from two vendors are not a trend and they are not a head-to-head with anything.

The useful posture here is to treat Vercel's line the way you would treat any vendor telling you what is popular on its own platform. Note it, then go run the eval that actually decides your architecture.

Diogo Almeida

@CompleteSkeptic

X

Jevtember