System One · Decision Models

Decision Models: AI for fast flows

A new category of models that return typed decisions with calibrated probabilities — 100× faster, 400× cheaper than LLMs.

The engineer who helped create the technology behind ChatGPT is now betting that the future of artificial intelligence is not chatbots.

What are Decision Models?

Daniel Kahneman split thinking into System 1 (fast, automatic) and System 2 (slow, deliberative). Applied to AI: frontier LLMs behave like System 2 — they write, token by token. Decision Models (also called System One models or classifier models) behave like System 1 for software: they return a typed decision plus calibrated probabilities, not prose. Typical shapes: choice, score, boolean. They sit in front of expensive or risky calls and leave generation to an LLM when you actually need one (dual-process). They can still pick the wrong option — restricted output is not perfect judgment.

The catalog tracks hosted Jev plus open peers (Kev, Nimble, Laya, NanoJev, jevlike, …). Learn the category, then compare on the quadrant.

Learn the category →

Featured Decision Model

TypeSafe AI

Jev

Waitlist

First public System One model — typed decisions with calibrated probabilities for software, not chat.

Latency
70–500ms
Pricing
$0.042/M in · out free
Types
choice · score · boolean
Full details