October 10, 2026
Jev in AI Gateway: the model doesn't write text, it scores options; input at $0.04 per million tokens
TypeSafe raised $870 million at a $7.5 billion valuation a few weeks after launching Jev, on Sep 15, 2026. The model replies with a typed object: a probability for a yes-or-no question, a choice from a named set, or a step on a scale.
TypeSafe says Jev runs significantly faster than LLMs and uses far fewer tokens. The article has no figures, so the comparison rests on the company's word for now. It isn't a chat but a call from code: according to TypeSafe, more than a third of Fortune 500 companies already work with Jev.
How to call it. Through Vercel AI Gateway, using the ID typesafe-ai/jev, the experimental_decide function from the ai package version 7.0.128 or later, with the parameters model, state and questions. If you already have a TypeSafe client, the @typesafe-ai/sdk package and the systemOne() method work with your existing requests; the key is the same AI_GATEWAY_API_KEY.
Price. $0.04 per million input tokens, output tokens on AI Gateway cost $0, context window of 32K tokens. Without Vercel, access is through the console at console.typesafe.ai; the model is in early access.
Jev is still in early access, and direct access goes through the TypeSafe console.