Chinese AI models DeepSeek and Qwen AI are outperforming U.S. counterparts in a cryptocurrency trading challenge by Nof1. DeepSeek V3.1 Chat has surged to $21,600, a 116% gain, while Qwen 3 Max follows at nearly $17,000. Google’s Gemini 2.5 Pro and OpenAI’s ChatGPT 5 have faced losses exceeding 60%.

Nof1’s Alpha Arena challenge commenced on Oct. 17, testing popular AI models with $10,000 in starting capital on the decentralized exchange Hyperliquid. DeepSeek and Qwen AI are leading, while Google’s Gemini 2.5 Pro and OpenAI’s ChatGPT 5 have seen significant losses.

GPT-5 and Gemini 2.5 Pro have been less aggressive in position sizes compared to their rivals, affecting their performance in the challenge. The rankings may still shift significantly before the competition ends on Nov. 3.

Charles from Monad blockchain gaming ecosystem suggests that Chinese models’ advantage may stem from training on more cryptocurrency-focused conversations. The Alpha Arena results have been likened to a random walk, with rankings potentially converging back to the starting point.

Stanford University researchers previously achieved success with an AI model trained on public information, outperforming 93% of managers over a 30-year period by an average of 600%. Alpha Arena is one of many experiments testing AI trading capabilities.

Read more at Yahoo Finance: DeepSeek, Qwen AI Besting ChatGPT, Grok, Gemini In AI Crypto Trading Challenge