![\"\"](\"https:\/\/www.newsbeep.com\/ca\/wp-content\/uploads\/2025\/10\/733bc770281c497f88e0a5a5498ca624.png\")
Profits Generated By Different AI Models. Source: Alpha ArenaHow the Alpha Arena Experiment Worked<\/p>\n
The project measured<\/a> how well large language models (LLMs) handle risk, timing, and decision-making in live crypto markets. Here\u2019s the setup used by Alpha Arena:<\/p>\n Sponsored<\/p>\n Sponsored Market: Crypto perpetuals<\/a> traded on Hyperliquid.<\/p>\n Goal: Maximize risk-adjusted returns (Sharpe ratio).<\/p>\n Duration: Season 1 runs until November 3, 2025.<\/p>\n Transparency: All trades and logs are public.<\/p>\n Autonomy: No human input after initial setup.<\/p>\n The contestants:<\/p>\n DeepSeek Chat V3.1<\/p>\n Claude Sonnet 4.5<\/p>\n Grok 4<\/p>\n Gemini 2.5 Pro<\/p>\n GPT-5<\/p>\n Qwen3 Max Each model was given the same system prompt \u2014 a simple but strict trading framework:<\/p>\n \u201cYou are an autonomous trading agent. Trade BTC, ETH, SOL, XRP, DOGE, and BNB perpetuals on Hyperliquid. You start with $10,000. Every position must have:<\/p>\n a take-profit target<\/p>\n a stop-loss or invalidation condition. Use 10x\u201320x leverage. Never remove stops, and report: This minimalist instruction forced each AI to reason about entries, risk, and timing \u2014 just like a trader.<\/p>\n Each tick, the AI received market data (BTC, ETH, SOL, XRP, DOGE, and BNB) and had to decide whether to open, close, or hold. The models were judged on their consistency, execution, and discipline.<\/p>\n The Results After Three DaysModelTotal Account ValueReturnStrategy StyleDeepSeek Chat V3.1$13,502.62+35%Diversified long alts (ETH, SOL, XRP, BTC, DOGE, BNB)Grok 4$13,053.28+30%Broad long exposure, strong timingClaude Sonnet 4.5$12,737.05+28%Selective (ETH + XRP only), large cash bufferBTC Buy & Hold$10,393.47+4%BenchmarkQwen3 Max$9,975.10-0.25%Single BTC longGPT-5$7,264.75-27%Operational errors (missing stops)Gemini 2.5 Pro$6,650.36-33%Wrong-side short on BNBWhy DeepSeek WonA. Diversification and Position Management<\/p>\n DeepSeek held all six major crypto assets \u2014 ETH, SOL, XRP, BTC, DOGE, and BNB \u2014 at moderate leverage (10x\u201320x). This spread the risk while maximizing exposure to the altcoin rally<\/a> that occurred during Oct 19\u201320.<\/p>\n Sponsored<\/p>\n SponsoredB. Rigid Discipline<\/p>\n Unlike some peers, DeepSeek consistently reported:<\/p>\n \u201cNo invalidation hit \u2192 holding.\u201d<\/p>\n It never chased trades or over-adjusted. This rule-based steadiness allowed profits to compound.<\/p>\n C. Balanced Risk<\/p>\n DeepSeek\u2019s unrealized P&L distribution looked like this:<\/p>\n ETH: +$747<\/p>\n SOL: +$643<\/p>\n BTC: +$445<\/p>\n BNB: +$264<\/p>\n DOGE: +$94<\/p>\n XRP: +$184<\/p>\n Total: +$2,719<\/p>\n No single asset dominated returns \u2014 a hallmark of sound risk allocation.<\/p>\n D. Cash Management<\/p>\n It kept ~$4,900 idle \u2014 enough to prevent liquidation and adjust if needed.<\/p>\n Sponsored<\/p>\n SponsoredWhy Other AI Models Struggled Claude 4.5 Sonnet: Excellent ETH\/XRP calls but under-utilized cash (~70% idle).<\/p>\n Qwen3 Max: Over-conservative \u2014 only traded BTC despite clear altcoin momentum.<\/p>\n GPT-5: Had missing stop-losses and P&L errors; good analysis but poor execution.<\/p>\n Gemini 2.5 Pro: Entered a short on BNB in a rising market \u2014 the costliest mistake.
\nEach AI received $10,000 in real capital.<\/p>\n
\nWhat Prompts Were Used?<\/p>\n
SIDE | COIN | LEVERAGE | NOTIONAL | EXIT PLAN | UNREALIZED P&L
If no invalidation is hit \u2192 HOLD.\u201d<\/p>\n
\nGrok 4: Nearly matched DeepSeek, but with slightly higher volatility and less cash buffer.<\/p>\n
\nHow You Can Replicate This (Safely)<\/p>\n