Thirty-five percent of paid teams on Vercel’s platform were using Jev within two weeks of its launch. That’s not a projected figure or a waitlist number. That’s active adoption, and it’s the fastest any model has reached that threshold in Vercel’s history.
According to TechCrunch, Vercel published the data on Thursday, showing Jev reaching nearly double the two-week adoption rate of Sonnet 5.5, which was previously the most popular model on the platform. For a model that only entered limited release on September 15, that’s a sharp result, and it gives TypeSafe AI its first concrete proof point after weeks of mostly anecdotal enthusiasm.
Jev is not trying to be a general-purpose model. It’s a decision model, meaning it’s built specifically to choose between a small set of predetermined outputs rather than generate open-ended responses. That constraint is the point. By narrowing what the model can do, TypeSafe AI made it significantly cheaper and faster to run than alternatives like Anthropic’s Sonnet or OpenAI’s frontier offerings. If you need to classify large volumes of data quickly, that tradeoff makes a lot of sense.
Vercel itself put that logic into practice, replacing GPT-6 Luna with Jev for classifying harmful commands. That’s a real production use case from a company that runs one of the most widely used deployment platforms for AI-powered applications. It’s also the kind of signal that tends to carry weight with other engineering teams evaluating where to spend their inference budget.
The data comes from Vercel’s AI gateway, a unified API layer comparable to OpenRouter. It’s worth being clear about what that means for interpretation. The gateway is more likely to undercount usage from OpenAI and Anthropic, since those labs have their own direct APIs that many teams use without going through a third-party layer. So Jev’s lead on this particular leaderboard reflects adoption within Vercel’s ecosystem, not the broader market. Still, within that context, the signal is real.
The buzz around Jev was already hard to miss before this data dropped. OpenAI and Amazon both responded with competing decision-model offerings in the weeks after Jev’s release, which is about as clear a validation signal as you can get. But competing products and social media momentum are not the same as usage data, and until now the good press lacked hard numbers to back it up.
The Vercel report also tracks a broader shift worth paying attention to. Open source models now account for more than 70% of tokens processed through the gateway and nearly 25% of total spend. That’s a meaningful share of actual workloads moving away from proprietary frontier models. But here’s the part that cuts against a simple narrative: spending on frontier labs has grown tenfold even as their share of overall spend has shrunk. The market is expanding fast enough that both categories are growing in absolute terms.
That matters because it challenges the assumption that open source and proprietary models are locked in a zero-sum competition. The data suggests teams are using both, often for different things. Lightweight, specialized models like Jev handle high-volume classification tasks. Frontier models handle reasoning, generation, and anything that needs more depth. So the winner in this environment probably isn’t the model that does everything, but the one that does a specific job better than anyone else at a price that makes sense to run at scale.
- Jev hit 35% adoption among paid Vercel teams within two weeks of limited release
- Nearly double the two-week reach of the second most popular model, Sonnet 5.5
- Designed for classification tasks with predetermined outputs, not open-ended generation
- Vercel replaced GPT-6 Luna with Jev for harmful command classification
- Open source models now represent over 70% of tokens on Vercel’s gateway
- Frontier lab spending has grown tenfold even as its share of total spend has decreased
For developers and founders evaluating their AI stack, Jev’s early numbers make it worth a closer look, especially for any workload that involves classifying data at volume. TypeSafe AI has gone from launch buzz to documented adoption faster than anyone else has managed on Vercel’s platform. That’s a real milestone, even if the broader market picture is still coming into focus.
