Evidence 1: LMSYS Leaderboard
Evidence 2: OpenAI Simple Evals Benchmark
Evidence 3: Analyzing the Sentiment of the OpenAI Community
Evidence 4: Costs
Evidence 5: Speed
Evidence 6: Model Capabilities and Conceptual Differences
Evidence 7: My Personal Opinion
Clearly, GPT-4o wins. It was able to process my whole 160-page dissertation and provided well-thought out research gaps that could be used for scientific progress. GPT-4 is not capable of that.
To learn more about this mind-boggling experiment, check out my blog and YouTube video: