Jev introduces a new shape of LLM - System One, aka Decision Models

TypeSafe AI has recently released Jev, a model that defines a new category they call “System One,” which some describe as “decision models.” Unlike traditional large language models, Jev takes text or semi‑structured data and returns numeric values representing probabilities, scores, or binary decisions, along with a confidence level. The API supports three query types: yes/no questions (called “Noul”), multiple‑choice selections, and scoring within a defined range. Each query is processed in parallel, so response time remains consistent even when many questions are sent simultaneously.

One of Jev’s most notable features is its cost: you pay only for input tokens at $0.042 per million tokens, cheaper than OpenAI’s GPT‑5 Nano. Its speed and low price make it useful for classification, spam detection, tagging, prioritization, and reordering search results. However, because it outputs only a single floating‑point number, the model behaves like a black box, raising concerns about interpretability and potential bias in its decisions.

This development is significant because it offers an economically efficient tool with automated decision‑making capabilities, expanding options for developers seeking affordable, task‑oriented AI solutions. It also highlights the need to carefully assess transparency and fairness before deploying such systems in critical contexts.

Read the original article on Simon Willison

This summary is an informational synthesis produced by dataqbs.com. All rights to the original content belong to its author and the cited media outlet. We act solely as curators of technology news and claim no authorship.