Skip to content

Compare · release 2026-09-18

Jev vs Qwen3.8 Flash

Jev vs Qwen3.8 Flash, scored against human labels. Decision Score: noul 69.0 vs 62.4, choice 67.8 vs 62.4 and score 9.2 vs −1.4. Release 2026-09-18.

Jev is a System One model: it answers with probabilities and writes no text. Qwen3.8 Flash is an LLM, answering the same questions through an adapter. How Jev differs from an LLM

noul on PubMedQA

Jev Qwen3.8 Flash
Decision Score69.095% range 59.8 to 76.562.495% range 50.9 to 72.5
Ranktied for rank 1 of 8rank 2 of 8
Accuracy91.3%89.7%
Calibration gap5.0 pts2.5 pts
Hand-off at 95%86%—
$ per 1k decisions$0.029$0.086
Speed p95653 ms3.5 s

Question by question (right = most of 5 runs right), out of 300: both right on 260, only Jev on 14, only Qwen3.8 Flash on 8, neither on 18. Same usual pick on 93%. See where they differ

choice on Banking77

Jev Qwen3.8 Flash
Decision Score67.895% range 61.3 to 74.562.495% range 55.9 to 68.9
Ranktied for rank 2 of 8tied for rank 4 of 8
Accuracy79.7%76.8%
Calibration gap9.8 pts9.5 pts
Hand-off at 95%54%—
$ per 1k decisions$0.043$0.12
Speed p95693 ms3.1 s

Question by question (right = most of 5 runs right), out of 300: both right on 218, only Jev on 20, only Qwen3.8 Flash on 11, neither on 51. Same usual pick on 88%. See where they differ

score on HelpSteer2 helpfulness

Jev Qwen3.8 Flash
Decision Score9.295% range −4.4 to 21.0−1.495% range −15.4 to 11.5
Ranktied for rank 1 of 8tied for rank 3 of 8
Accuracy41.3%36.1%
Calibration gap19.7 pts25.8 pts
Hand-off at 95%——
$ per 1k decisions$0.036$0.13
Speed p95670 ms2.6 s

Question by question (right = most of 5 runs right), out of 300: both right on 86, only Jev on 38, only Qwen3.8 Flash on 23, neither on 153. Same usual pick on 68%. See where they differ

Jev · Qwen3.8 Flash · all models · how we score · board.json