Jev reaches 35% of Vercel's paid teams in two weeks, the fastest model adoption the gateway has measured

Vercel published gateway data on Thursday showing that Jev, the lightweight decision model from TypeSafe AI, was being used by 35% of the platform's paid teams within two weeks of its September 15 launch. That is nearly double the reach of the second most popular model on the gateway, Sonnet 5.5, and the fastest first-two-weeks adoption of any model in Vercel's measurements, according to TechCrunch.

The number is the first hard usage data behind a model that had, for a month, mostly buzz.

What Jev actually is

Jev is not a general assistant. It is a specialized model for a narrow job: choosing between a few predetermined outputs. The restricted output space is the whole design. With less to generate, the model runs cheaper and faster than general alternatives like Sonnet, which makes it a fit for classifying large quantities of data quickly. TechCrunch notes the approach has since inspired competing offerings from OpenAI and Amazon, among others, so "decision model" is becoming a category, not a one-off.

Vercel is not a neutral reference point either. The company was an early adopter and uses Jev internally to replace GPT-6 Luna for classifying harmful commands.

The measurement has a known blind spot

The data comes from Vercel's AI gateway, a unified API service in the OpenRouter mold. As TechCrunch points out, that is a necessarily limited view: traffic that goes straight to a lab's own API, particularly OpenAI's and Anthropic's, never touches the gateway. Any adoption ranking built on this data undercounts the majors, which is relevant here because the runner-up, Sonnet 5.5, is an Anthropic model.

That does not kill the signal. A model at 35% of paid teams through a single gateway, ahead of everything else routed through that gateway, is a usage pattern, not a rumor. The gateway data converts the good vibes around Jev into at least one independently observable number.

The two numbers in the report worth keeping

Beyond Jev, TechCrunch highlights two figures that say more about the market:

• Open source models now account for more than 70% of the tokens processed through Vercel's gateway, while representing nearly 25% of total spend.
• Spending on frontier labs has grown tenfold over the period, even as their share of overall gateway spend has shrunk.

TechCrunch's read is that frontier models and open source alternatives may turn out to be complementary rather than competitive: the open models absorb the high-volume, lower-stakes tokens, while the frontier bills grow tenfold on the work that still needs them.

What this is worth for routing decisions

If you have a high-volume classification or routing workload, the Jev data point is a reason to look at specialized decision models before defaulting to a frontier general model. The economics follow directly from the design: cheaper and faster inference on a task that only needs a choice from a fixed set. The caveat is scope. The numbers come from one gateway's paid teams, through one company's measurement, and the most comparable competitors may be invisible to it.

Treat the 35% as the strongest evidence yet that the decision-model niche is real, and re-run the arithmetic on your own traffic before swapping models.