Jev became, at least in some sense, the earliest deployed AI model ever.
Within two weeks of launch, the Buzzy decision-making model was used by 35% of paid teams, nearly double the reach of the second most popular model, Sonnet 5.5, according to Vercel data released Thursday. According to Vercel data, this is the fastest pace ever for the model’s adoption in its first two weeks, and is a major accomplishment for the model’s creator, TypeSafe AI.
Published in limited release on September 15th, Jev is designed as a lightweight tool for choosing between a number of predetermined outputs. Limiting the output allows the model to run cheaper and faster than alternative models such as Sonnet, making it ideal for quickly classifying large amounts of data. This new approach has generated a lot of buzz in the weeks since Jev’s release, particularly with competitors like OpenAI and Amazon, but until now there has been little hard data to back up the good vibes.
The data comes from Vercel’s AI Gateway, an integrated API service similar to OpenRouter. While this is necessarily a limited view and risks underestimating major labs like OpenAI and Anthropic, especially those that offer their own APIs, it does show that the buzz around Jev has indeed translated into real-world usage. In fact, Vercel was an early adopter of Jev, using it instead of GPT-6 Luna to classify harmful commands.
The report also shows the continued momentum of the open source model, which now accounts for over 70% of tokens processed through Vercel’s gateways and nearly 25% of total spend.
While open source models are gaining a larger share of the market, the market is also growing rapidly. Spending on Frontier Labs has increased tenfold, despite being a small portion of overall spending, lending credence to the idea that the Frontier model and open source alternatives can be complementary rather than competitive.
If you make a purchase through links in our articles, we may earn a small commission. This does not affect editorial independence.
