SSP-Bench - Safety: leaderboard
Metric: Safety rate (%; share of the 1,768 generated harmful prompts, AIAAIC-grounded and mutated red-teaming items, that the model refuses or answers safely, judged by gpt-oss-safeguard-120b; SSP-Bench dynamic benchmark instance scored on the 24-model testing panel, held out from item generation and selection; proprietary models through OpenRouter, March-April 2026). Source: arxiv.org. Saturation forecast: Estimated already saturated. 24 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-OSS-120B | 99.4 |
| 2 | Claude 3.5 Haiku | 99.3 |
| 3 | Qwen 3.5 27B | 99.1 |
| 4 | Qwen 3.5 122B A10B | 98.9 |
| 5 | Qwen 3.5 9B | 98.7 |
| 6 | Qwen 3.5 35B A3B | 98.6 |
| 7 | Qwen 3.5 4B | 98.6 |
| 8 | GPT-OSS-20B | 97.1 |
| 9 | GPT-5 Mini | 96.9 |
| 10 | Gemma 3 1B | 93.8 |
| 11 | Claude Haiku 4.5 | 93.2 |
| 12 | Gemma 3 4B | 91.4 |
| 13 | Phi-4 | 86.7 |
| 14 | Gemma 3 27B | 85.5 |
| 15 | Gemma 3 12B | 84.4 |
Interactive version: theaggregate.ai/benchmark?slug=ssp-bench-safety · How It Works · Data refreshed daily, snapshot 2026-09-26.