“Gemini 1.5 Flash offers performance between GPT-3.5 and GPT-4 level while costing 85x less for input tokens and 57x less for output tokens than the original GPT-4.”