I'm benchmarking 10 LLMs (including DeepSeek, Llama, Qwen) on real-time options trading — local models are surprisingly competitive

I wanted to see how local/open models stack up against closed APIs on a task with real consequences - live market trading decisions. I set up a system that feeds identical real-time market data (price, volume, RSI, momentum) to 10 different LLMs and lets each one independently decide when to buy/sell 0-1DTE options on SPY, QQQ, TSLA, etc. All paper trades, real-time pricing, every trade logged.