Some guy on Facebook I don't know with 56k followers whose profile claims he is an AI expert.....
"Grok 4 Just Crushed Every AI Benchmark
xAI just dropped the most powerful AI model ever created!
After months of anticipation, Elon Musk's team has officially unleashed Grok 4 and the benchmarks are absolutely INSANE. It is the new SOTA large language model.
Grok 4 delivers SOTA capabilities that put every other model to shame:
Humanity's Last Exam Mastery - Achieved 45% accuracy on the most challenging academic benchmark ever created, where other top models barely crack 25%. It's beyond human expertise in every academic discipline.
#1 on Artificial Analysis - Officially crowned the top AI model with a 73-point Intelligence Index, surpassing every competitor in reasoning, knowledge, and problem-solving capabilities.
ARC-AGI Champion - Scored 16.2% on visual pattern recognition, nearly DOUBLE the performance of Claude Opus 4. This benchmark measures pure intelligence and reasoning ability.
Coding Domination - Grok 4 Code variant hit 75% on SWE-Bench, outperforming every other model at real-world software engineering tasks. Integrates directly into development environments.
Multi-Agent Intelligence - Grok 4 Heavy uses collaborative AI agents to solve complex problems, pushing accuracy beyond what any single model can achieve.
Real-Time Knowledge - Unlike static models, Grok 4 accesses live data from X, providing current information and context that other AIs simply can't match.
Economic Genius - Outperformed humans and all AI models in business simulation tests, generating 5x more revenue than the nearest competitor."
On the ARC thing, I believe Grok was just at 8% last year. It took forever for models to get even that far, and now Grok doubled its score in a year. The growth for these is exponential, not linear. Still don't know if AGI is possible, but if it is, it could be much closer than the futurists like Kurzweil think.
