SemIf placed second with a score of 74.6 on a new performance metric for Jev class models released Sept. 19. The benchmark, called JevBench, provides the first standardized evaluation for the category and shows the original Jev model by Typesafe AI in the lead at 75.3.

The 0.7 point difference between the two top models indicates a narrow margin between the current leading tools in the ecosystem. Full results for the benchmark were released on a dedicated page alongside the formal introduction of the JevBench system.