Access Restricted for EU Residents
You are attempting to access a website operated by an entity not regulated in the EU. Products and services on this website do not comply with EU laws or ESMA investor-protection standards.
As an EU resident, you cannot proceed to the offshore website.
Please continue on the EU-regulated website to ensure full regulatory protection.
Tuesday Nov 4 2025 11:10
2 min
In a surprising turn of events, two Chinese artificial intelligence chatbots, QWEN3 MAX and DeepSeek, have surpassed some of the world's most advanced models, including OpenAI’s ChatGPT, in an autonomous cryptocurrency trading competition that concluded on Tuesday.
The budget-friendly AI models, QWEN3 MAX and DeepSeek, secured the first and second positions, respectively, in the trading challenge, outperforming competitors with higher profiles and significantly larger budgets. According to data aggregator CoinGlass, QWEN3 was the only AI chatbot to generate positive returns, achieving a total profit of $751 with a 7.5% return rate. In contrast, all other AI bots concluded the competition with losses.
OpenAI’s ChatGPT trailed behind, incurring a substantial 57% loss, reducing its initial investment of $10,000 to a mere $4,272 by the competition's end.
To secure the victory in the trading competition, QWEN3 employed a 20x leveraged long position on Bitcoin (BTC). The AI models, as of Wednesday, exclusively open positions. QWEN 3 initiated the leveraged bet when Bitcoin was trading at $104,556. CoinGlass data indicates that the position faces potential liquidation if BTC falls below $100,630.
Prior to the competition's conclusion, QWEN 3 primarily maintained leveraged long positions on Bitcoin, Ether (ETH), and Dogecoin (DOGE).
The competition's unexpected results emphasize that even the most extensively funded AI models may still lack the real-time capabilities necessary for successful cryptocurrency trading. Despite OpenAI allocating $5.7 billion to research and development initiatives in the first half of 2025 alone, according to Reuters, ChatGPT finished last.
While QWEN3’s budget remains undisclosed, machine learning engineer Aakarshit Srivastava estimates that the model’s training may have cost between $10 million and $20 million. DeepSeek, despite being developed at a total training cost of $5.3 million, according to the model’s technical paper, secured second place.
Alpha Arena’s competition commenced with each bot receiving $200 in starting capital, which was subsequently increased to $10,000 per model, with trades executed on the decentralized exchange Hyperliquid.
Risk Warning: This article is provided for informational purposes only and does not constitute investment advice, investment research, or a recommendation to trade. The views expressed are those of the author and do not necessarily reflect the position of Markets.com. When considering shares, indices, forex (foreign exchange), and commodities for trading and price predictions, remember that trading CFDs involves a significant degree of risk and may not be suitable for all investors. Leveraged products can result in capital loss. Past performance is not indicative of future results. Before trading, ensure you fully understand the risks involved and consider your investment objectives and level of experience. Cryptocurrency CFD trading restrictions may apply depending on jurisdiction.
Latest
View all
Monday, 21 September 2026
6 min
Monday, 21 September 2026
10 min
Monday, 21 September 2026
10 min
EN
