Benchmarks & research
Interfaze’s Lev runs on your own machine and scores the answers you supply, rather than writing a reply.
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
Interfaze reports 68.9% on all thirteen S1Bench subsets, compared with Jev’s 76.1%. Its open model adapts Qwen3.5-4B and accepts TypeSafe-style requests. The Lev here is Interfaze’s project: TypeSafe’s Jev and Xvectorio’s Lev operations tool are separate projects.
Lev runs on your own computer. Send it some text, a question and the answers it may choose from; the response gives each answer a score, with no composed reply. An application that normally sends questions to TypeSafe can send the same kind of request to Lev instead.