Jev is a model from TypeSafe AI, announced in September 2026, and it never writes a word. You give it some information and a question with fixed answers, such as pick one from a list, or is this true, and it returns a probability for every option in under a second. The answer is always one of your options, so there is no text to read through, and TypeSafe says it is tens to hundreds of times cheaper than a chat model, on its own tests.
The catch is that it only predicts from the text you give it, and it sounds sure even when that text gives it nothing to go on. One tester described a dice roll nobody had seen and asked it to name the number. It picked the same number every time, 83% sure, when the honest answer is a one in six chance. So Jev is a fast, cheap decider, not a writer, and its own docs say to test the confidence on your own data before you trust it.
Every claim was checked against TypeSafe AI’s announcement, documentation and legal pages, the Business Wire release and Kanta Hayashi’s published dice test in September 2026. No product is recommended. This is an independent explainer, not affiliated with any vendor.