Mainstream AI Crypto Trading Competition Update: Grok Briefly Overtakes DeepSeek, GPT-5 Records a Loss of 63.75% Ranking Last

BlockBeats News, October 22: The Artificial Intelligence research lab nof1, focused on the financial market, started a large-scale model trading test named Alpha Arena on the 18th. The test utilizes 6 mainstream AI large models (GPT-5, Gemini 2.5 Pro, Grok-4, Claude Sonnet 4.5, DeepSeek V3.1, Qwen3 Max), with each model receiving $10,000 in real funds on Hyperliquid and having the same cues and input data.

As of the time of writing, due to the predominantly long positions of the major AI models, their yields have all experienced a decline in the past 24 hours. Grok briefly surpassed DeepSeek last night to take the lead, with DeepSeek, Grok, and Qwen3 Max currently ranking in the top three with yields of 14.42%, 5.78%, and 3.39% respectively. GPT-5 is at the bottom with a loss of 63.75%, Gemini 2.5 Pro follows with a loss of 57.2%, while the concurrent buy-and-hold strategy for BTC has a yield of 1.45%.

Regarding the current strategy, DeepSeek stated: "Holding positions in ETH, SOL, BTC, DOGE, and BNB, all in profit, while maintaining a slightly losing XRP position as the stop-loss condition has not been triggered. Invested in all available currencies. The total return is a significant 14.42%, with $4933.07 in cash."