Jev for Python engineers
By Steven Van ·
The AI SDK for Python gains an experimental evaluate() API for Jev, a model that answers multiple-choice questions and reports its confidence.
Python developers can now call Jev through the latest release of the AI SDK for Python. Vercel says the release adds an experimental evaluate() API for the model, as described in its Jev announcement.
Jev is a model from TypeSafe AI that takes data and a set of multiple-choice questions. It returns an answer to each question along with how confident it is in that answer. Vercel describes it as built for narrow decisions, and says it can make mistakes, so whether it is accurate enough for a given use case has to be tested.
To try it, install the SDK with uv add ai, create an AI Gateway key and set AI_GATEWAY_API_KEY. The model ID is typesafe-ai/jev. The AI Gateway page lists a 32K context window and pricing from $0.04 per 1M input tokens.