# Decision model benchmark: how accuracy falls as reasoning gets deeper

> 10 models on 3,600 ProofWriter reasoning problems. Jev leads the open-world task at 83.8%, ahead of GLiDE, Oct 2 at 82.9%. Accuracy by proof depth, paired comparisons, repeatability and latency.

10 models on 3,600 ProofWriter reasoning problems. Jev leads the open-world task at 83.8%, ahead of GLiDE, Oct 2 at 82.9%. Accuracy by proof depth, paired comparisons, repeatability and latency.
