TypeSafe AI's Jev now available on AI Gateway

TypeSafe AI’s Jev, a probabilistic decision model, is now available on Vercel’s AI Gateway. Unlike traditional language models that generate text token by token, Jev evaluates all declared questions in parallel and returns typed answers—Choice, Score, or Boolean—along with probabilities. This approach eliminates unnecessary text generation and simplifies automation by routing clear cases for immediate action while flagging uncertain ones for review.

TypeSafe AI reports that Jev demonstrated performance gains of up to 193.6x faster execution and 444.6x lower costs compared to large language models in workflow evaluations. The model is exposed through AI SDK 7’s experimental evaluate API, which supports structured decision-making with typed outputs such as Choice for option selection, Score for grading, and Boolean for probability estimation.

To use Jev, developers must install AI SDK 7.0.105 or later and call the model via the identifier typesafe-ai/jev. The system accepts a map of named decisions to process against a given state, returning results with preserved question IDs and confidence metrics in the providerMetadata.typesafe.confidence field.

Jev supports Zero Data Retention and No Training, configurable per request. Evaluation calls are logged, count toward usage budgets, and can be integrated into custom reporting. The feature is accessible alongside other Gateway provider options within the providerOptions.gateway object, with documentation available for evaluation models on AI Gateway.