Using large models for contract trading is a joke - GPT keeps losing without turning back - DeepSeek had a major pullback from the highs, following the peak trades led to huge losses - Risk-adjusted return: not a single large model can outperform holding Bitcoin, and this doesn’t even account for the time and money costs of running large models Why large models fail: using large models without fine-tuning or feeding specialized data for contract trading is like asking a language teacher to solve advanced math problems.