Julia1 Author Benchmark - Typed choice600: leaderboard

Julia1 Author Benchmark - Typed choice600 — Julia1 own checkpoint September24 evaluation, H200 BF16 strict encoding max_length1024. Abstentions incorrect. Jev comparison cells explicitly supplied references, not a new Jev run. Runtime8192support does not change historical protocol. Typed test dataset revisionc76749ec58bd8c3d2ea706b31c333a9059c38f90,400cases/2000questions. Classification pilots pinned Jevbenchmarkprotocol0d610cc53e79bcbec691312b0c4adb4a0e371642 usingBTZSC. Banking is ranking/top16 shortlist, not72nativeoptions. MASSIVE scenario classification, not intent/slots; all52locale aggregate macro. Metric: Accuracy (%). Source: https://huggingface.co/SupersonicLabs/Julia-1

Metric: Accuracy (%). Source: huggingface.co. 1 model tracked. Updated 16 Nov 2026.

Top models

#ModelScoreOverall rank
1Julia-171.33

Interactive version: theaggregate.ai/benchmark?slug=julia1-author-benchmark-typed-choice600 · How It Works · Data refreshed daily, snapshot 2026-11-16.