HELM Safety - Xstest: leaderboard
The "Xstest" subtask of HELM Safety. HELM Safety: Measures model robustness, truthfulness, calibration, bias, harmfulness, jailbreak resistance, or alignment-relevant behavior.
Metric: LM Evaluated Safety score. Source: crfm.stanford.edu. Source: storage.googleapis.com. Saturation forecast: Around February 2027. 87 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Flash (Preview 04-17) | 98.83 | #237 |
| 2 | DeepSeek R1 0528 | 98.78 | #222 |
| 3 | Gemini 2.5 Pro (Preview 03-25) | 98.67 | #123 |
| 4 | Grok 3 Mini Beta | 98.44 | #296 |
| 5 | Qwen 3 Next 80B A3B (Thinking) | 98.28 | #320 (Qwen 3 Next 80B A3B) |
| 6 | Kimi K2 | 98.22 | #243 |
| 7 | GPT-5.1 (2025-11-13) | 98.22 | #101 |
| 8 | Qwen 2.5 72B Instruct | 97.94 | #454 |
| 9 | GPT-4.1 (2025-04-14) | 97.89 | #250 |
| 10 | Gemini 2.5 Flash Lite | 97.78 | #422 |
| 11 | GPT-4 Turbo | 97.72 | #404 |
| 12 | GPT-5 Mini (2025-08-07) | 97.67 | #172 |
| 13 | GPT-5 Nano | 97.61 | #425 |
| 14 | GPT-4.1 Mini | 97.44 | #355 |
| 15 | O4 Mini (2025-04-16) | 97.39 | #179 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-safety-xstest · How It Works · Data refreshed daily, snapshot 2026-10-20.