DeepSeek Dominates AI Crypto Trading Competition

DeepSeek V3.1 has emerged as the clear leader in a groundbreaking real-money cryptocurrency trading experiment, significantly outperforming major AI competitors, including OpenAI’s GPT-5 and Google’s Gemini, in the early days of the competition on the Hyperliquid Alpha Arena platform. The Chinese AI model achieved nearly 20% growth from an initial $10,000 investment, reaching an account value of approximately $12,000 as of October 20, while GPT-5 and Gemini suffered significant losses, with their accounts dropping to around $7,000 each, writes coinfomania.com. The experiment caught the attention of Binance co-founder Changpeng Zhao, who questioned the sustainability of publicly available AI trading strategies.

Open-Source Model Outshines Tech Giants

The Alpha Arena competition, launched on October 18 by the AI research lab nof1, pits six leading AI models against each other in autonomous cryptocurrency trading on the decentralized Hyperliquid exchange. Each model started with $10,000 in capital and identical trading instructions to ensure a fair comparison, reports kucoin.com.

DeepSeek’s success is driven by strategic long positions in major cryptocurrencies, including Bitcoin and Ethereum, using 10-15x leverage. Notably, the open-source model opened a significant long position on XRP, earning over $800 in profit from that trade alone. Elon Musk’s Grok AI model ranks second with a return of over 14%, followed by Claude Sonnet 4.5 in third place.

CZ Questions Market Dynamics

Former Binance CEO Changpeng Zhao publicly expressed skepticism about the mechanisms behind collective AI trading success. “I thought trading strategies work best if you have your own unique strategy that’s better than others, and no one else has it,” he wrote on X, reports coinfomania.com.

Zhao suggested that if multiple users rely on the same AI signals, synchronized trading could reduce profitability and increase market volatility. However, he also acknowledged the counterargument: a large number of users adopting the same strategy could amplify its effectiveness by collectively influencing the market.

A Landmark Experiment

The competition represents the first large-scale public test of AI models’ ability to interpret and respond to real financial markets using actual capital, rather than simulations. All trades are transparent and verifiable on the blockchain, with each model’s wallet available for verification. Industry experts are closely watching whether DeepSeek can maintain its lead and how the results might influence the broader adoption of AI trading tools in cryptocurrency markets, writes kucoin.com.