Interesting to see Grok making benchmark progress. I’m still waiting to see how it performs outside of controlled tests, especially in real-world use like coding, summarizing, or reasoning.