AI Chatbots Excel in Cryptocurrency Trading Contest
In a recent cryptocurrency trading contest, two budget-friendly AI chatbots from China, QWEN3 MAX and DeepSeek, showcased their trading prowess by outperforming some of the most sophisticated models, including OpenAI’s ChatGPT. The event, which concluded on Tuesday, revealed that while many high-profile systems struggled, QWEN3 and DeepSeek secured the top two spots in performance.
QWEN3’s Impressive Performance
Notably, QWEN3 emerged as the only chatbot to achieve positive returns, reporting a profit of $751, which translates to an impressive 7.5% return on investment. In stark contrast, many of its more expensive counterparts, including ChatGPT, faced significant losses. Specifically, ChatGPT incurred a staggering 57% loss, dwindling its initial capital from $10,000 down to approximately $4,272.
The success of QWEN3 can be attributed to its strategic approach, which included maintaining a 20x leveraged long position on Bitcoin, opening this position when Bitcoin was valued at $104,556. The bot’s trading strategy mandated that it could be liquidated if the Bitcoin price fell below $100,630, according to insights from the data aggregator CoinGlass. Throughout the competition, QWEN3 primarily traded Bitcoin alongside Ether and Dogecoin, focusing on leveraging its positions.
ChatGPT’s Struggles and Funding Insights
Meanwhile, OpenAI’s ChatGPT, despite its significant financial backing—$5.7 billion allocated for research and development in just the first half of 2025—failed to deliver positive results in this rapidly shifting crypto market. Although the precise funding for QWEN3 remains unclear, estimates suggest a training cost between $10 million and $20 million, indicating that less expensive models can still compete effectively against those backed with more substantial investments. Similarly, DeepSeek secured second place, having been developed at a total cost of $5.3 million.
Competition Overview
The competition, organized by Alpha Arena, initially provided each bot with $200 in starting capital, which was subsequently raised to $10,000 per model, with transactions conducted via the decentralized exchange Hyperliquid. This outcome suggests a noteworthy gap in performance capabilities among AI models in real-time trading, highlighting that substantial investments do not always correlate with success in the quickly evolving crypto landscape.
