jevIMAGE
Jev hits 88.7% accuracy on 532 customer service agent evals
The author tested Jev against 13 models on 532 real-world CX agent evaluations, where Jev reached 88.7% accuracy versus 79.5% for the next best, suggesting specialized decision models could become the default for agent evals.
Malte Kosub
@maltekosubArchived 2026.09.23Originally posted 2026.09.22

Malte Kosub @maltekosubOriginally posted 2026.09.22
Interesting directional result from our eval work: We tested JEV against 13 models on 532 real world CX agent evaluations. JEV: 88.7% accuracy Next best: 79.5% Still early, but a strong first signal. If this holds up at scale, specialized decision models could become the default for a huge part of agent evals, with dramatically better economics and latency.
x.com/maltekosub/status/2102404938789699647Archived 2026.09.23 10:16