Chinese AI Models Lead in Crypto Trading Contest
Two Chinese artificial intelligence chatbots, QWEN3 MAX and DeepSeek, ranked first and second in a recent cryptocurrency trading competition, surpassing international peers including OpenAIs ChatGPT. The event concluded Tuesday, highlighting shifts in AI trading capabilities.
Contest Details and Performance
The autonomous trading challenge was tracked by CoinGlass, with each AI bot initially allocated $200, later increased to $10,000 per model. Trades were executed on the Hyperliquid platform. The competition measured profitability and return rates among participating bots.
- QWEN3 MAX was the only AI to report positive returns, achieving profits of $751 and a 7.5% return rate.
- All other AI bots, including DeepSeek and ChatGPT, closed in negative territory.
- OpenAIs ChatGPT was last, ending with a 57% loss and a closing balance of $4,272 from its $10,000 starting capital.
Strategies and Technical Insights
QWEN3 MAX primarily used leveraged long positions, focusing on Bitcoin (BTC), Ether (ETH), and Dogecoin (DOGE). At the end of the competition, QWEN3 MAX held a 20x long position on Bitcoin. This position was opened when Bitcoin was priced at $104,556 and risked liquidation if BTC dropped below $100,630, according to CoinGlass data.
Development Investment and Real-Time Performance
The competition results drew attention to performance gaps among well-known AI models. OpenAI invested $5.7 billion in research and development in the first half of 2025, according to Reuters. QWEN3s training cost is estimated between $10 million and $20 million, while DeepSeek was reportedly developed with a total training cost of $5.3 million, as outlined in its technical paper.
Industry Implications
Alpha Arena, the organizer, began the competition on October 18. The event's outcome raises questions over the effectiveness of present AI models in real-time crypto trading environments. While some models with significant funding failed to deliver positive returns, lesser-known bots outperformed in practical trading scenarios. The results may spur further innovation in AI-driven crypto strategies and model development.
Related content
Comments
