Grok 4.5 Testing Results: How SpaceXAI’s New Model Performs on Real Professional Work

We’ve evaluated Grok 4.5 on Snorkel’s GDPval+ dataset, Snorkel’s expert-created dataset of professional workplace reasoning tasks from across the economy. To compare performance against other frontier models, we ran the evaluation alongside GPT 5.5 and Claude Opus 4.8. Overall, Grok 4.5 demonstrated the strongest overall performance.