Translated data: Microsoft's paper reveals that GPT-3.5 has only 2 billion parameters, a significant discrepancy from the official claims. The small model CodeFusion performs exceptionally well with just 75 million parameters. Amidst the controversy, the AI community is fiercely competitive, with ChatGPT and ClaudeAI vying for dominance. Meanwhile, GPT-4 has demonstrated astonishing capabilities in reviewing manuscripts, and even Terence Tao has used GPT-4 to solve mathematical problems.