#AGI 「AI 模型炒幣實測」實驗中，六款前沿大型語言模型各獲 $10K 操作加密衍生品，結果五個虧損，GPT-5 更慘蝕 $5,679，僅 Qwen3 MAX 小虧 $652 勉強勝出。這場名為 Alpha Arena 的比賽突顯：AI 在高波動市場表現仍「極具人性」。
Why it matters: The experiment exposes the current limits of LLMs in real-world financial applications—especially in volatile, high-risk environments.
The big picture: While AI excels at pattern recognition, trading requires a blend of judgment, context, and adaptability—traits even the most advanced models still lack.
📌 一杯咖啡價錢連接 Web3 世界 https://patreon.com/wanszezit
Full article https://www.reuters.com/technology/alpha-arena-ai-models-crypto-trading-losses-2025-11-07/