Jev has achieved the fastest adoption rate ever recorded for an AI model, according to at least some metrics. Data released on Thursday reveals that within just two weeks of launch, this buzzy decision model was already being utilized by 35% of paid teams—nearly twice the reach of Sonnet 5, the second most popular model.

EMBED_PLACEHOLDER_0

According to the data, this represents the quickest adoption any model has seen in its first two weeks, marking a significant achievement for Typesafe AI, the model's creators.

Released in limited form on September 15, Jev functions as a lightweight tool for selecting between a small set of predetermined outputs. By constraining the outputs, the model can operate more cheaply and quickly than competitors like Sonnet, making it perfect for rapidly classifying large volumes of data. This novel approach has generated considerable excitement since Jev's release—spurring competing products from OpenAI and Amazon, among others—though until now, there has been scant concrete evidence to support the enthusiasm.

The data originates from Vercel's AI gateway, a unified API service comparable to OpenRouter. This provides a necessarily limited perspective—particularly prone to undercounting major labs like OpenAI and Anthropic that provide their own APIs—but it demonstrates that the excitement around Jev genuinely converted into real usage. In fact, Vercel was an early adopter of Jev, employing it to replace GPT-6 Luna for classifying harmful commands.

The report also indicates sustained momentum for open-source models, which now represent over 70% of the tokens processed through Vercel's gateway and nearly 25% of total spending. While open-source models are claiming a larger market share, the market itself is expanding rapidly. Spending on frontier labs has increased ten-fold, even as it constitutes a smaller portion of overall spending, lending support to the notion that frontier models and open-source alternatives may prove to be complementary rather than competitive.