Key Info

OpenRouter tested Jev, a decision model from Typesafe, against popular LLMs on a judging/classification task using Ori Eval. Jev was more than 5x faster than the next-fastest model, matched the accuracy of leading classifiers, and was the second cheapest of the five — just behind Qwen3.8 Flash.

Highlights

  • Jev was over 5x faster than the next-fastest LLM on the task.
  • Accuracy matched popular classification LLMs; all five models landed within a handful of cases of one another.
  • Cost ranked second lowest, just behind Qwen3.8 Flash and below three other LLMs.
  • The results support Typesafe's thesis that decision models can be fast, accurate, and cheap.