Laya Release - MASSIVE51 Macro Accuracy: leaderboard

CPU 51-language sweep, 100 cases per language, 20-option intent questions (random 0.050) 5100 Per-language rows (la-11) optional. Metric: macro accuracy / macro F1 / macro ECE (as served) / count.

Metric: macro accuracy / macro F1 / macro ECE (as served) / count. Source: raw.githubusercontent.com. 1 model tracked. Updated 14 Nov 2026.

Top models

#ModelScoreOverall rank
1Laya (421M)0.2269

Interactive version: theaggregate.ai/benchmark?slug=laya-release-massive51-macro-accuracy · How It Works · Data refreshed daily, snapshot 2026-11-16.