Ollama running llama3:70b I only get 1token per second with a 4090 but all the models smaller than that hallucinate frequently.