TypeSafe AI released Jev, an AI model that delivers fast judgments inside software instead of writing text, on September 16, 2026. It is the company's first major update since its founding as a startup.
Jev is priced at $0.042 per million input tokens, significantly lower than typical AI models, and offers responses in 70 to 500 milliseconds, many times faster than current language models. That compares with standard AI models that take several seconds to generate responses.
Jev is built to deliver narrow judgments and probabilities inside other programs, targeting use cases like customer service and sales. Availability begins with a waitlist, initially for developers.
"Sorting requests instead of answering customers," said Diogo Almeida, co-founder and CEO. "Jev allows developers to define questions and possible answers, and the model scores the options."
The announcement follows TypeSafe AI's founding as a startup. The company frames the significance of Jev as a shift toward more efficient and targeted AI use within software.
TypeSafe AI did not say how Jev's judgments will compare to human accuracy, and the model's reliability remains something companies will have to test for their own tasks. The company plans to release more details on Jev's performance in the coming months.
Source: thedecoder
