SimpleQA (with search)
Search · 2024-10-30
SimpleQA's short fact-seeking questions answered by a model with a web-search tool or search-agent framework rather than closed-book. Search access lifts scores far above closed-book SimpleQA, so these runs are kept apart from the closed-book row; the search framework is a run-level setting. Higher is better.
Top models (higher is better)
| Model | Score |
|---|---|
| DeepSeek-V3.2-Exp | 97.1 |
| DeepSeek-V3.1-Terminus | 96.8 |
| DeepSeek-V3.1 | 93.4 |
| DeepSeek-R1-0528 | 92.3 |