DeepSeek V3.1 вырвался в лидеры криптотрейдинга, обогнав Qwen и OpenAI Translated headline: DeepSeek V3.1 Takes the Lead in Crypto Trading, Outpacing Qwen and OpenAI

The Chinese AI model DeepSeek V3.1 achieved a 120% return within nine days during the cryptocurrency trading competition Alpha Arena organized by the nof1 lab. It generated $22,031 from an initial capital of $10,000.

The contest is set to continue until November 3rd. Six AI models were each allocated $10,000, provided with the same instructions and market data. Their objective is to maximize profits by trading Bitcoin, Ethereum, and Dogecoin.

The Qwen 3 Max neural network from Alibaba Group is currently in second place with an 80% return, raising its balance to $18,014. For a period, Qwen was ahead of DeepSeek but eventually fell behind.

Models from OpenAI and Google DeepMind incurred substantial losses. GPT-5 suffered a 61.5% capital decrease, leaving it with $3,845. Gemini 2.5 Pro reported a similar outcome, with a loss of 59% ($4,077).

Other participants displayed more modest performances. Grok 4 from xAI recorded a 9% gain ($10,911), while Claude 4.5 Sonnet from Anthropic achieved a 21% increase ($12,122).

It’s worth noting that DeepSeek has introduced a new multimodal AI capable of processing large and complex documents while utilizing significantly fewer tokens.