FACTS Benchmark Suite - Grounding, Multimodal and Search
General QA
Incomplete FACTS suite average over Grounding v2, Multimodal v1 and Search v2, with Parametric absent. This exact three-component mean is preserved for provenance and excluded from capability factors because it is not comparable with the complete suite or the separate three-component average without Search.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.6 Sol | 52.4 |