OpenAI’s ChatGPT lost 63% of its money in a two-week crypto trading competition hosted by Nof1, finishing last of six major language models (LLMs) according to Protos.
AI Bots test crypto trading skills
The ‘Alpha Arena’ competition, which ended on Monday, tasked six leading AI systems to trade digital assets using identical prompts and limited data sets.
ChatGPT, Google’s Gemini van Alphabet (NASDAQ:GOOGL), X’s Grok and Anthropic’s Claude Sonnet all finished in the red.
On the other hand Alibabas (NYSE:BABA) Qwen3 Max topped the leaderboard with earnings of $2,232, followed by DeepSeek, which gained $489.
The rest saw steep losses: ChatGPT lost $6,267, Gemini fell $5,671, Grok fell $4,531 and Claude fell $3,081, from their starting balance of $10,000.
Trading costs erode AI performance
No1 said profits were “dominated by trading fees in early runs” as agents overtraded and took small profits that offset the fees.
Gemini recorded 238 transactions, while Claude made only 38. For all six models, win rates ranged between 25% and 30%.
Qwen3 Max had the highest total fees at $1,654, but still outperformed its competitors thanks to its disciplined trade selection.
The consistent profitability of the Chinese model stands in stark contrast to ChatGPT’s heavy losses, highlighting the divergent risk behavior among LLMs under identical circumstances.
Organizers call it a stress test for AI
Founder of Nof1 Jay Azhang described the event as a controlled stress test for generative AI systems.
“LLMs don’t really handle numerical time series data well, but that’s the whole context we gave them,” Azhang said, noting that each model had to deal with “strict rules and limited context windows.”
He added that each AI exhibited a unique “investing personality,” indicating predictable trends in the way language models approach markets.
Azhang plans to organize a new round of the competition with refined clues and greater statistical accuracy.
Why it matters
The competition shows that language models can sound confident, but still fail when real money is at stake.
LLMs processed the same charts and data, but their results varied, just like human traders with different risk habits.
Qwen3 Max succeeded not by speed, but by avoiding overtrading, proving that discipline trumps prediction.
The loss of ChatGPT highlights that market execution is more important than ideas or stories.
Investors are learning that AI can help analyze markets, but cannot replace strategy or risk management.
Read next:
Image: Shutterstock
#ChatGPT #lost #trade #crypto #Chinese #healthy #profit