Low-cost AI Beats ChatGPT in Crypto Trading
In a competition of autonomous trading chatbots organized by Alpha Arena, an unexpected twist occurred. Two Chinese artificial intelligence models — QWEN3 MAX and DeepSeek — managed to outperform the famous ChatGPT from OpenAI, despite costing only a fraction of its development budget.
Cheaper model, better results
According to CoinGlass data, QWEN3 was the only chatbot that managed to make a profit in crypto trading. From an initial investment of $10,000, it earned $751 in profit, representing a 7.5% return. The second place went to another Chinese model, DeepSeek, while all others, including ChatGPT, ended up in the red.
At the bottom of the ranking was ChatGPT, which lost 57% of its investment — from the original $10,000, only $4,272 remained. OpenAI, meanwhile, invested $5.7 billion in model development during just the first half of 2025, according to Reuters.
By contrast, QWEN3’s development reportedly cost between $10 and $20 million, while DeepSeek was built with a budget of around $5.3 million.
Small budget, big performance
The competition began on October 18 with an initial capital of $200 per bot, later increased to $10,000. All trades took place on the decentralized exchange Hyperliquid.
The results show that in crypto trading, success depends not only on budget or computing power but also on a model’s ability to react to the market in real time — something many advanced Western AI systems still struggle with.
What it means for the crypto world
The success of Chinese models serves as both a warning and an inspiration. It demonstrates that efficient, well-trained AI can outperform even the most expensive technologies if properly tuned to its environment. As artificial intelligence increasingly enters financial markets, China is emerging as a strong player — technologically advanced and economically competitive.
Sources:
https://www.coinglass.com/ai/hyperliquid
https://arks0001.medium.com/qwen-qwq-32b-whats-different-c3a53c400e33
