Airoboros 33b GPT4 1.2 merged with kaiokendev's 33b SuperHOT 8k LoRA, without quant. (Full FP16 model)