
Nof1’s October 18 Alpha Arena trial tested six AI models on Hyperliquid with $10,000 each, revealing diverse outcomes in real-world market trading performance.
On October 18, AI lab nof1 ran the Alpha Arena trading test with six major AI models, each allocated $10,000 on Hyperliquid. DeepSeek achieved a 14.42% return, Grok followed at 5.78%, and Qwen3 Max posted 3.39%. GPT-5 incurred a 63.75% loss, while a Bitcoin buy-and-hold strategy yielded 1.45%. The results illustrate varying trading capabilities among AI models under the same conditions.