After seeing the text version of gpt4o I am quite sure that the next™ model will reach 95% of gpt4 performance with just 20% of the cost!