A recent experiment by Nof1 lab evaluated the performance of six major AI models in live cryptocurrency trading on the Hyperliquid platform. Each model, including ChatGPT, Gemini, and QWEN3, was allocated $10,000 and provided with identical market data to trade cryptocurrencies like BTC, ETH, SOL, BNB, DOGE, and XRP. Over a two-week period, four models incurred losses, with ChatGPT losing $6,267, Gemini $5,671, Grok $4,531, and Claude Sonnet $3,081. In contrast, DeepSeek and QWEN3 managed to generate profits of $489 and $2,232, respectively. The experiment aimed to explore AI's potential in crypto trading but underscored challenges such as high trading costs and decision-making difficulties in volatile markets.