In a new experiment called the Alpha Arena, top AI models are competing to trade cryptocurrencies using real money. So far, Elon Musk’s Grok, DeepSeek, and Claude Sonnet 4.5 are leading, each earning more than 25% returns. Interestingly, DeepSeek’s portfolio is up 36% over the weekend.
Other well-known models, including OpenAI’s GPT-5 and Google’s Gemini 2.5 Pro, are not performing as well. Both have seen heavy losses of more than 28% since the contest began on October 17.
Our new benchmark has the top 6 AI models trading real capital
Grok4 is winning so far. It was short and then flipped to long, timing the bottom perfectly
It’s up >500% in 1 day pic.twitter.com/k6bOZzLGkF
— Jay A (@jay_azhang) October 10, 2025
The competition, created by Nof1, an AI research firm, runs until November 3. Each model started with $10,000 in real capital to trade crypto perpetual contracts on the Hyperliquid exchange. These contracts allow traders to bet on price changes for coins like Bitcoin, Dogecoin, and Solana, without expiry dates.
How the Alpha Arena Works
Each AI trades fully on its own. It decides what to buy or sell, how much to risk, and when to exit. All trades are public to ensure transparency.
@grok you’re wild pic.twitter.com/YSyh13gtzH
— Jay A (@jay_azhang) October 10, 2025
According to Nof1 founder Jay Azhang, it’s no surprise that Grok and DeepSeek are leading. “It usually ends up between Grok and DeepSeek,” he said, adding that GPT sometimes performs better in longer tests.
GPT-5, in contrast, took a very cautious approach. It traded rarely, which limited its profits but also kept it from major losses. Meanwhile, Claude Sonnet 4.5 ranked third, showing steady performance.
Prediction market for outcome of Alpha Arena is about to go live on @Polymarket
— Jay A (@jay_azhang) October 20, 2025
“DeepSeek and Grok seem to have better contextual awareness of market microstructure. Grok in particular has made money in 100% of the past 5 rounds,” Azhan wrote.
AI Trading Still Has Risks
The contest highlights both the promise and danger of AI in trading. Models like Grok show that general-purpose AIs can learn markets fast. But failures from Gemini and others reveal the risks; AI can act unpredictably or make costly errors.
DeepSeek’s portfolio is up 36% over the weekend
How about yours? pic.twitter.com/fRqowHYfc5
— Jay A (@jay_azhang) October 20, 2025
While this technology could change how markets work, experts say it’s still early days. For now, the Alpha Arena is proving that even the smartest AIs can struggle to master the crypto markets.