# System one models

A System One model is a model built to make a decision rather than write text. You give it
input, such as a support ticket, a JSON object, or an agent's last tool result, plus the
questions you want answered and the answers allowed. It returns one of those answers with a
probability, in a single pass, in well under a second. The name borrows Kahneman's split
between fast, intuitive System 1 thinking and slow, deliberate System 2 reasoning. TypeSafe
AI coined it in September 2026 when it launched Jev, and open models now use the label too.

The point is the probability. A model whose "90% sure" answers are right about 90% of the
time lets software set a threshold: act on its own above it, and hand the case to a person or
a reasoning model below it. An LLM that states its own confidence can't be used that way,
because the confidence it reports doesn't track how often it's right.

Most of what an agent does between reasoning steps is this kind of decision: which route,
which tool, whether a request is a refund, whether an output is safe to send. A System One
model takes those off the LLM. It does not explain its answers and it cannot generate, so
anything that needs reasoning or writing still goes to an LLM.

## Questions answered under this concept

- [Are there open-source System One models?](https://tessary.ai/answers/system-one-models/are-there-open-source-system-one-models)
- [Can a decision model work as a guardrail on LLM output before it's sent?](https://tessary.ai/answers/system-one-models/can-a-decision-model-guardrail-llm-output-before-its-sent)
- [Can a decision model route requests between a cheap LLM and an expensive one?](https://tessary.ai/answers/system-one-models/can-a-decision-model-route-requests-between-a-cheap-llm-and-an-expensive-one)
- [Does calibration hold when my production traffic drifts from what the model was trained on?](https://tessary.ai/answers/system-one-models/does-calibration-hold-when-production-traffic-drifts)
- [How do I check whether a model's confidence is calibrated on my own traffic?](https://tessary.ai/answers/system-one-models/how-do-i-check-if-model-confidence-is-calibrated-on-my-traffic)
- [How do I pick the confidence threshold for letting a decision run automatically?](https://tessary.ai/answers/system-one-models/how-do-i-pick-a-confidence-threshold-for-a-decision)
- [How is a System One model different from an LLM with structured outputs?](https://tessary.ai/answers/system-one-models/system-one-model-vs-llm-structured-outputs)
- [What is a System One model?](https://tessary.ai/answers/system-one-models/what-is-a-system-one-model)
- [What's the difference between Jev and AWS's Strands Decider?](https://tessary.ai/answers/system-one-models/whats-the-difference-between-jev-and-aws-strands-decider)

---

Source: https://tessary.ai/answers/system-one-models
All concepts: https://tessary.ai/answers
From Tessary, agent reliability for AI agents in production: https://tessary.ai