ChatGPT-5 performed worst among major AI models in a recent trading tournament, losing nearly 60% of its starting $10,000 portfolio and finishing with around $4,000. The competition tested AI systems on real-time market trading and risk management over several weeks.
Only one model achieved positive results – Qwen3-Max – which gained roughly 10%, ending near $11,000. DeepSeek 3.1 also posted a minor loss, closing at about $9,300. The results highlighted significant performance gaps between leading AI systems under live market conditions.
Online commentators suggested that ChatGPT-5’s underperformance may stem from its widespread public usage, arguing that its general-purpose tuning limits its ability to adapt to fast-changing market data compared with specialized financial AIs.