Grok 4.5 achieved an Elo score of 1328 on the Website Arena, marking a 25-rank jump from its previous highest-performing model. The model, trained with Cursor data, now performs in the same band as Claude Opus 4.6 (Thinking) for real-world website-building tasks.
xAI's Grok 4.5 demonstrates competitive coding capabilities against frontier models like Claude Opus, intensifying the race for practical agentic AI applications.