FACTS Benchmark Suite - Grounding, Multimodal and Parametric
General QA
Incomplete FACTS suite average over Grounding v2, Multimodal v1 and Parametric v1, with Search absent. This exact three-component mean is preserved for provenance and excluded from capability factors because it is not comparable with the complete four-component suite.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.5 | 66.7 |
| Gemini 2.5 Pro | 61.5 |
| GPT-5 | 56.5 |
| Gemini 3.1 Flash-Lite Preview | 51.4 |
| GPT-5.2 | 48.5 |
| Gemini 2.5 Flash | 47.2 |
| Grok 4 | 46.3 |
| GPT-4.1 | 45.8 |
| Opus 4.7 | 45.5 |
| GPT-5.1 | 45.0 |
| o3 | 44.4 |
| Opus 4.5 | 43.9 |
| Opus 4.6 | 41.5 |
| Sonnet 4.5 | 40.6 |
| Opus 4.1 | 40.4 |
| GPT-5 Mini | 38.6 |
| Sonnet 4.6 | 37.0 |
| Sonnet 4 | 35.0 |
| Gemma 3 27B IT | 29.1 |
| O4 Mini | 28.1 |
| Grok 4 Fast | 25.5 |