Jev reaches 13% of paid Vercel teams within 24 hours of launch
Jev makes decisions instead of writing sentences, returning results in 70–500ms with free output, and is already available on Vercel AI Gateway and as a Claude Code skill for content moderation and ticket routing.
An ex-OpenAI researcher shipped an AI that is 444x cheaper than Claude and can't write a single sentence. It's called Jev, and it launched on September 15. Within 24 hours, 13% of paid teams on Vercel's AI Gateway had plugged it in, more than 2x any model launch before it. Here's the trick: Claude and GPT write 1 token at a time, while Jev doesn't write at all, it decides. It gives yes or no, A, B or C, a score, or a pick from up to 255 options, each with a probability. The whole output looks like this: "decision": "YES", "label": "spam", "confidence": 0.97 It answers in 70 to 500 milliseconds and costs $0.042 per 1M input tokens, with output free. On narrow decision tasks, it runs up to 194x faster than an LLM. Now look at your API bill. Is this email spam, is this comment safe, is this lead hot, did the agent's last step pass? Most AI calls in production ask for a 1-word answer, and most teams pay a frontier model to write that word. Jev fits content moderation, lead scoring, sentiment analysis, ticket routing and quality checks between agent steps. There are 3 ways to plug it in today. It's already in the model list on Vercel AI Gateway. You can grab a key from the TypeSafe API. Or you can install the Jev skill in Claude Code, and Claude hands every yes/no call to Jev mid-session. The company raised a $40M seed led by DCVC, and Cloudflare, LangChain and Langfuse added Jev within 3 days. Claude does the thinking, Jev makes the call. The cheapest model in your stack is the one that won't talk.