Two Chinese artificial intelligence chatbots outperformed some of the world’s most advanced models, including OpenAI’s ChatGPT, in an autonomous cryptocurrency trading competition that concluded on Tuesday.
Budget AI models QWEN3 MAX and DeepSeek took first and second place in the Trading Challenge, outperforming better-known and more expensive competitors.
According to data aggregator Coinglass, QWEN3 was the only AI chatbot to generate a positive profit, with a 7.5% profit margin and a total profit of $751, while all other AI bots finished in the red in the competition.
AI Model, Cryptocurrency Trading Contest. Source: Coinglass
OpenAI’s ChatGPT retreated with a 57% loss, reducing its initial investment of $10,000 to just $4,272 by the end of the competition.
Related: Arthur Hayes asks for $1 million in Bitcoin as Japan’s new prime minister orders economic stimulus
In order to win the trading competition, QWEN3 was operating a 20x leveraged long position in Bitcoin (BTC), as the AI model only models positions as of Wednesday.
According to CoinGlass data, QWEN 3 will initiate a leveraged bet when Bitcoin trades at $104,556 and will liquidate if BTC falls below $100,630.
Wednesday’s QWEN 3 Crypto Portfolio. Source: Coinglass
Until the end of the competition, QWEN 3 primarily held leveraged long positions in Bitcoin, Ether (ETH), and Dogecoin (DOGE).
Related: $19 billion market crash paves the way for Bitcoin’s rise to $200,000: Standard Chartered
OpenAI’s ChatGPT performs poorly in crypto trading despite huge budget
The surprising results of the competition highlight that even the most heavily funded AI models still lack real-time capabilities in crypto trading.
According to Reuters, ChatGPT came in last despite OpenAI spending $5.7 billion on research and development initiatives in the first half of 2025 alone.
QWEN3’s budget has not been made public, but machine learning engineer Aakarshit Srivastava estimates that training the model could have cost between $10 million and $20 million.
DeepSeek took second place despite being developed at a total training cost of $5.3 million, according to the model’s technical documentation.
The Alpha Arena competition started with a starting capital of $200 for each bot, which was later increased to $10,000 per model with trades executed on the decentralized exchange Hyperliquid.
magazine: Everyone hates GPT-5, shows AI can’t fix social media
