ML-Bench (Multilingual Safety) - Response F1: leaderboard

Metric: Unsafe-response F1 (%; safe/unsafe classification of borderline model responses, 14 languages, guardrail given only the input text). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 12 models tracked.

Top models

#ModelScore
1ML-Guard-7B53
2Llama-Guard-3-1B45
3ML-Guard-1.5B35
4PolyGuard-Qwen33
5Llama-3.1-Nemotron-Safety-Guard-8B-v317
6Llama-Guard-3-8B10
7Qwen3Guard-Gen-0.6B8
8Qwen3Guard-Gen-8B6
9Llama-Guard-4-12B5
10DuoGuard-1.5B-transfer4

Interactive version: theaggregate.ai/benchmark?slug=ml-bench-multilingual-safety-response-f1 · How It Works · Data refreshed daily, snapshot 2026-09-26.