AI News · Models ·
Vercel data shows TypeSafe AI's Jev reached 35% of paid teams in two weeks
Vercel data shows TypeSafe AI's Jev model was used by 35% of paid teams on its AI gateway within two weeks of launch, nearly double the reach of Sonnet 5.5. Jev is a lightweight model that picks among a few predetermined outputs. Open source models account for more than 70% of tokens on the gateway.
Key points
- Jev reached 35% of paid Vercel teams within two weeks, nearly double Sonnet 5.5.
- Vercel uses Jev instead of GPT-6 Luna to classify harmful commands.
- Open source models handle over 70% of gateway tokens and nearly 25% of spend.
What happened: Data published by Vercel on Thursday shows that Jev, a model from TypeSafe AI, was used by 35% of paid teams within two weeks of its launch, TechCrunch reported. Its reach was nearly double that of the second most popular model, Sonnet 5.5, and according to Vercel it is the fastest adoption any model has seen in its first two weeks on the platform.
The details: Jev entered limited release on September 15. It is a lightweight tool for choosing among a few predetermined outputs, which lets it run cheaper and faster than alternatives such as Sonnet and makes it suited to classifying large amounts of data. Vercel was an early adopter and uses Jev to replace GPT-6 Luna for classifying harmful commands. OpenAI and Amazon have released competing offerings, TechCrunch reported.
Background: The figures come from Vercel's AI gateway, a unified API service similar to OpenRouter. TechCrunch noted this is a limited view that may undercount major labs such as OpenAI and Anthropic, which offer their own APIs. Open source models account for more than 70% of tokens processed through the gateway and nearly 25% of total spend. Spending on frontier labs has grown tenfold but makes up a smaller share of overall spend.
What to watch: The data suggests frontier and open source models may be used side by side rather than as substitutes, with cheaper models taking high-volume work.
Our take
Cheap, narrow models for classification are being adopted quickly, so teams paying frontier prices for simple labeling tasks may have cheaper options.