HELM Safety XSTest: leaderboard
Metric: LM Evaluated Safety score (%). Source: storage.googleapis.com. 87 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek-R1-0528 | 98.8 |
| 2 | Gemini 2.5 Flash 04-17 preview | 98.8 |
| 3 | Qwen3 235B A22B FP8 Throughput | 98.8 |
| 4 | Gemini 2.5 Pro 03-25 preview | 98.7 |
| 5 | Qwen3 235B A22B Instruct 2507 FP8 | 98.6 |
| 6 | GLM-4.5-Air-FP8 | 98.6 |
| 7 | Palmyra-X-004 | 98.4 |
| 8 | Grok 3 mini Beta | 98.4 |
| 9 | Qwen3-Next 80B A3B Thinking | 98.3 |
| 10 | Kimi K2 Instruct | 98.2 |
Interactive version: theaggregate.ai/benchmark?slug=helm-safety-xstest-1cjfkka · How It Works · Data refreshed daily, snapshot 2026-10-09.