Are DeepSeek's publicly models themselves used in the trading operations of the HFT? I'd assume not, I feel even the architecture of models must be significantly different when comparing what R1 does versus what I'd assume trading models would do.