Jev AI Model: turn text into structured decisions through an API

About Jev AI Model

Jev AI Model is built for a narrow but common need in software: turning a piece of text into a decision your code can act on. Instead of a chat reply, you get structured answers to questions you define. The homepage example takes a support message ("I was charged twice this morning") and returns a choice (route to the billing queue), a yes or no with a probability (96 percent likely urgent) and a score (urgency 2.8 out of 3).

A playground lets you try this before writing code. It comes with about two dozen example tasks, including support routing, agent routing, citation checking, content moderation, lead scoring and LLM guardrails. You can load an example without signing in, edit the input and questions, then sign in to run it. Several models are available, including Jev 1.13, Liquid d1, Upstage Solar Decide, Kev 4B and the Laya models for English and other languages, with a benchmarks page for comparing them.

Pricing is pay as you go with no subscription. New accounts get 100 credits. Packs cost $10, $100 or $1,000, at $1 for 10,000 credits, and credits never expire. Successful calls start at 1 credit, charged on input tokens only; the pricing page lists rates per million input tokens by model, such as $0.06 for Jev and $0.03 for Span-01.