What was claimed
Jev, a model from TypeSafe AI released on September 15, was used by 35% of paid teams on Vercel's AI gateway within two weeks. Vercel calls it the fastest adoption the platform has seen, with reach nearly double that of the runner-up, Sonnet 5.5.
What is actually there
Jev is a narrow "decision model". It picks among a small set of predefined outputs instead of writing free text, which makes it cheap and fast for classifying large volumes of data. It is a tool for a specific chore, not a general assistant, so comparing its reach with a general model is comparing different jobs.
The number counts teams that used it, not how much they used it. A team that sent a hundred requests counts the same as one that sent a hundred million. Nothing in the report shows tokens, spend or whether teams stayed after trying it.
Why the headline is shaky
The data comes from a single gateway. Labs such as OpenAI and Anthropic also sell directly, so the gateway likely undercounts them, which flatters a newcomer. Vercel also adopted Jev early and used it to replace GPT-6 Luna for classifying harmful commands, so the company reporting the record has a stake in the model's success.
What the data does show
There is real interest in cheap, narrow models. Open-source models handle over 70% of tokens on the gateway, and frontier spending still grew tenfold. Useful context, but not proof that Jev beat the field.