jevbooks

← All projects

jev-eval

Shogo-nfrealmusic/jev-eval

Third-party check: Jev vs GPT-4o-mini and Claude Sonnet 4.5 on booking inquiry routing.

Evaluation & Benchmarking53%Runner-up: Classification & RoutingQuestion health measured

What it is

Compares Jev, TypeSafe AI's judgment-only model, against GPT-4o-mini and Claude Sonnet 4.5 under identical conditions on routing 60 synthetic booking inquiries for a photo-shoot service in Japan.

How it uses Jev

Jev is called via experimental_evaluate through the Vercel AI Gateway. The README does not specify which decisions Jev makes, what state it receives, or how its output is used in code.

Technique worth stealing

Measure Jev and LLMs under identical conditions with serial calls, warm-up, and raw response logging.

Try it

npm install; npx tsx src/smoke.ts; npm run measure; npm run score

View on GitHub

judged by Jevjev-1.13.0

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-centricno0.38
  • Shows a System One patternno0.09
  • Handles uncertaintyno0.04
  • Measuredno0.23
  • Runnableno0.14
  • Worth recommendingno0.26
  • Model replicano0.04
  • Problem scopescore on a 0–2 scale1.22
  • About Jevyes0.98

Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.