SimpleBench
Classic NLP · 2024-08-20
SimpleBench tests everyday commonsense reasoning with more than 200 multiple-choice questions spanning spatial and temporal reasoning, social intelligence, and adversarial trick questions. The official MCQ leaderboard reports accuracy averaged across five runs (AVG@5).
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5 | 81.9 |
| Claude Opus 5 | 80.6 |
| Gemini 3.1 Pro Preview | 79.6 |
| GPT-5.5 Pro | 76.9 |
| Gemini 3.5 Flash | 76.7 |
| Gemini 3 Pro Preview | 76.4 |
| GPT-5.6 Sol Pro | 71.7 |
| Qwen3.7-Max | 70.4 |
| Grok 4.5 | 70.0 |
| GPT-5.5 | 69.0 |
| Opus 4.6 | 67.6 |
| Opus 4.8 | 64.8 |
| GPT-5.6 Sol | 64.8 |
| Qwen3.6-Max-Preview | 63.0 |
| Gemini 2.5 Pro Preview 06-05 | 62.4 |
| Opus 4.5 | 62.0 |
| GPT-5 Pro | 61.6 |
| Gemini 3 Flash Preview | 61.1 |
| Kimi K3 | 60.7 |
| Sonnet 5 | 60.6 |
| Grok 4 | 60.5 |
| Opus 4.1 | 60.0 |
| Opus 4 | 58.8 |
| GLM-5.2 | 58.8 |
| Kimi K2.7 Code | 57.9 |
| GPT-5.2 Pro | 57.4 |
| GPT-5 | 56.7 |
| Grok 4.1 Fast | 56.0 |
| GLM-5.1 | 55.1 |
| Sonnet 4.5 | 54.3 |
| GPT-5.1 | 53.2 |
| GLM-5 | 53.2 |
| o3 | 53.1 |
| DeepSeek-V3.2-Speciale | 52.6 |
| Gemini 2.5 Pro Experimental 03-25 | 51.6 |
| Gemini 2.5 Pro Preview 03-25 | 51.6 |
| Inkling | 50.0 |
| GPT-5.6 Terra | 48.9 |
| GLM-4.7 | 47.7 |
| Kimi K2.5 | 46.8 |