A lot of people I talk to (in research, education & business) still don’t seem to know how much better GPT-4 is than GPT-3.5.
The difference is huge in basically every use case. You absolutely should be using GPT-4, which is free in Bing Creative Mode or else available for pay in ChatGPT Plus. Claude 2 is also good, though not quite GPT-4 level.
Here’s test score data as one example (the graph is from the Wharton School video, using data from OpenAI) , and the percentage is the percentage of human test takers beaten by the AI.