On November 4th, the highly anticipated AI large model real-time investment competition 'Alpha Arena' came to a close, with Alibaba's Qwen winning the final championship. The competition was initiated by the three organizations Nof1 on October 18th, gathering six of the world's top models including Qwen3-Max, DeepSeek v3.1, GPT-5, Gemini 2.5 Pro, Claude Sonnet 4.5, and Grok 4, each with an initial fund of $10,000, making autonomous decisions and trading in the real market without human intervention, determining the final champion based on profit and loss.
After 17 days, Alibaba's Qwen won the final championship with a return rate of over 20%. Qwen and DeepSeek, two Chinese models, were the only two models that made a profit, while all four top American models incurred losses, with GPT-5 losing over 60%, finishing at the bottom.
Jay Azhang, the founder of the event organizer Alpha Arena, has repeatedly praised Alibaba's Qwen model strategies and performance, congratulating Alibaba's Qwen for ultimately winning the championship. Industry experts have stated that the excellent performance of Alibaba's Qwen and DeepSeek in practical applications demonstrates the strong potential of Chinese models in solving real-world problems. The deep understanding of scenarios by AI will be key to the implementation of large models and future global AI competitions.