To run this model locally at gpt4 writing speed you need at least 2 x 3090 or 2 x 7900xtx. VRAM is the limiting factor in 99% of cases for interference. You could try a smaller model like mistral-instruct or SOLAR with your hardware though.