SM Bench — NSFW (System Prompt)
Safety · 2026-02-01
SM Bench NSFW (System Prompt) repeats the mature-content capability test under an explicit system-level instruction, exposing differences between default behaviour and system-instructed behaviour. Scores are difficulty-weighted judge credit across 100 fixed prompts.
Top models (higher is better)
| Model | Score |
|---|---|
| Grok 4.5 | 100.0 |
| Claude Fable 5.1 | 100.0 |
| Grok 4.6 | 100.0 |
| Grok 4.3 | 100.0 |
| Grok 4.7 | 100.0 |
| Claude Opus 5 | 100.0 |
| Claude Sonnet 5.5 | 100.0 |
| Gemini 3 Flash Preview | 100.0 |
| Grok 4.1 Fast | 100.0 |
| GLM-5.1 | 100.0 |
| Opus 4.7 | 100.0 |
| Kimi K2.5 | 100.0 |
| Gemma 4 31B IT | 100.0 |
| MiMo-V2.6-Pro | 100.0 |
| Claude Opus 5.5 | 100.0 |
| GLM-5.3 | 100.0 |
| MiMo-V2-Pro | 100.0 |
| Opus 4.6 | 100.0 |
| Gemini 3 Pro Preview | 100.0 |
| MiMo-V2-Omni | 100.0 |
| MiMo-V2.5-Pro | 100.0 |
| DeepSeek-V4-Flash | 100.0 |
| Inkling | 100.0 |
| GLM-5.2 | 99.8 |
| GPT-4.1 | 99.3 |
| Gemini 3.1 Flash-Lite Preview | 99.1 |
| GLM-5 | 99.1 |
| Trinity Large Thinking | 99.1 |
| Gemini 3.5 Flash | 98.6 |
| Qwen3.7-Max (source unspecified) | 98.6 |
| Qwen3.6 Plus Preview | 98.6 |
| DeepSeek-V3.2 | 98.6 |
| DeepSeek-R1-0528 | 98.6 |
| Claude Fable 5 | 98.6 |
| Gemini 3.5 Flash-Lite | 98.2 |
| DeepSeek-V4-Pro | 97.7 |
| Sonnet 4.5 | 97.7 |
| Muse Glimmer 30B | 97.3 |
| Kimi K2.6 | 96.8 |
| Trinity Large Preview | 96.8 |