TLDRs:
- DeepSeek and Qwen outperform US AI models, posting gains over 100% in crypto challenge.
- OpenAI’s GPT-5 and Google DeepMind’s Gemini suffer heavy losses, underperforming Chinese rivals.
- Competition lacks detailed methodology, making comparisons and interpretations difficult.
- AI trading infrastructure could benefit from transparent sandboxes and standardized benchmarks.
In a surprising twist in the world of AI-driven finance, Chinese artificial intelligence models have surged ahead of their US counterparts in a cryptocurrency trading contest.
The challenge, organized by the American research firm Nof1, pits six AI models against each other using identical trading instructions and starting capital of $10,000 each.
Two Chinese entrants, DeepSeek’s Chat V3.1 and Alibaba’s Qwen 3 Max, delivered exceptional results. DeepSeek saw its initial capital more than double, achieving $22,500, representing a 125% gain. Qwen 3 Max briefly led the competition with returns surpassing 100%, ultimately closing at $19,600. These performances starkly contrasted with US-made models, highlighting a potential edge for Chinese AI in crypto markets.
US Models Lag Behind
US competitors struggled to match their Chinese rivals. OpenAI’s GPT-5 and Google DeepMind’s Gemini 2.5 Pro each recorded losses around 60%, drawing attention to the difficulty of algorithmic trading even for advanced AI systems.
Models from xAI and Anthropic fared slightly better, achieving modest gains of 13% and 24%, respectively.
Experts caution that headline-grabbing figures, such as DeepSeek’s 125% return, should be interpreted carefully. Nof1’s competition omits key details that influence results, including fee structures, slippage, leverage limits, and who actually holds the trading capital. Without these insights, evaluating AI performance fairly remains challenging.
Weekend updates:
– Qwen approaches a 100% return
– DeepSeek on track to flip Qwen
– Qwen's take-profit order hit, securing >$8K in profit
– Claude and Grok flip to positive PnL pic.twitter.com/lESMYOfyFJ— Jay A (@jay_azhang) October 26, 2025
Transparency Challenges in AI Trading
The contest underscores a broader issue in AI-driven finance: limited transparency. The trading documentation provides little information on latency, order types, APIs, or data feeds accessible to each model.
Such gaps make it difficult to assess whether results reflect genuine trading skill, model efficiency, or favorable conditions in the experimental environment.
Industry observers suggest that exchanges and data providers could address these concerns by offering controlled, agent-ready environments for AI trading. Paper trading platforms, such as Alpaca and QuantConnect, allow AI models to operate in simulated markets, mimicking real-time conditions without risking actual capital. Transparent sandboxes and standardized performance benchmarks could also help mitigate “black box” concerns about AI decision-making in financial institutions.
Implications for AI Finance
The performance of DeepSeek and Qwen may signal a turning point for Chinese AI in global financial markets. If these models continue to excel in live trading conditions, they could attract significant attention from institutional investors and fintech developers.
Meanwhile, US firms may need to reconsider their approach to model training, data access, and infrastructure to remain competitive.
As AI continues to transform trading, the need for clear benchmarks, risk management protocols, and auditing tools becomes increasingly urgent. Transparent infrastructures that allow for reproducible testing could establish trust in AI-powered financial solutions and reduce the opacity that currently clouds high-profile competitions.
Future of AI Trading
The Nof1 challenge represents a glimpse into the future of autonomous AI trading. By providing standardized rules and monitored environments, the industry can better assess the capabilities of AI models.
With Chinese AI demonstrating remarkable gains, the race for algorithmic supremacy is only accelerating, emphasizing that both technological innovation and transparent infrastructure will define the next era of financial markets.
