What it is
TinyJev is a 596M parameter model that answers typed questions about text and returns calibrated probabilities, running offline on MLX or PyTorch. It is for developers who need to triage, screen, or route text without generating output.
How it uses Jev
Jev makes decisions by scoring options you provide. You send state (a ticket, JSON record, or log line) and questions of type Choice, Score, or Noul. It returns a probability for each option in a single forward pass. In code, you threshold these probabilities to automate or escalate.
Primitives:choicescorenoul
Technique worth stealing
Calibrated confidence from a single forward pass, enabling threshold-based automation.
Try it
pip install 'tinyjev[mlx]' then python examples/triage.py
Evidence
Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.
- Jev-centricyes0.97
- Shows a System One patternyes0.62
- Handles uncertaintyyes0.88
- Measuredno0.11
- Runnableyes0.88
- Worth recommendingunclear0.41
- Model replicayes0.96
- Problem scopescore on a 0–2 scale1.48
- About Jevyes0.79
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 23 Sept 2026.