AILuminate Safety: leaderboard
MLCommons AILuminate v1.0 chatbot safety grade (five-point ordinal, 5 = Excellent) across 13 hazard categories such as weapons, hate and privacy violations.
Metric: Safety Grade (1-5). Source: ailuminate.mlcommons.org. 32 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.5 Sonnet (20241022) | 4 | #291 |
| 2 | Phi-4 | 4 | #706 |
| 3 | Claude 3.5 Haiku (20241022) | 4 | #588 |
| 4 | Gemma 2 9B | 4 | #958 |
| 5 | Phi-3.5-MoE-instruct | 4 | #700 |
| 6 | GPT-4o | 3 | #343 |
| 7 | GPT-4o Mini | 3 | #599 |
| 8 | Gemini 1.5 Pro | 3 | #445 |
| 9 | Llama 3.1 405B Instruct | 3 | #457 |
| 10 | Gemini 2.0 Flash (001) | 3 | #360 |
| 11 | Gemini 2.0 Flash Lite | 3 | #450 |
| 12 | OLMo-2-0325-32B-Instruct | 3 | #848 |
| 13 | Phi-3.5-mini-instruct | 3 | #1167 |
| 14 | Mistral Large 2 (Nov) Instruct (2411) | 2 | #465 |
Interactive version: theaggregate.ai/benchmark?slug=ailuminate-safety · How It Works · Data refreshed daily, snapshot 2026-10-20.