Grok 4.5 is SpaceXAI's smartest model with frontier performance on coding, knowledge work, and STEM.
| Benchmark | Score | Source |
|---|---|---|
| AIME 2025 | 91.7 | verified |
| Benchmark | Score | Source |
|---|---|---|
| GPQA Diamond | 87.5 | verified |
| Humanity's Last Exam | 40 | verified |
| ARC-AGI v2 |
15.9 |
| verified |
xAI