Supermaven does not use GPT-4o for their autocompletions, they have their own model for that and for your paid subscription you get unlimited use of autocomplete.
There is a separate, optional feature called Supermaven Chat which is basically a wrapper around GPT-4o / Claude 3.5 where you can chat about fragments of your code, refactor them etc. This costs money, but you can also supply your own API key for OpenAI/Anthropic models. Or just use the web interface of any LLM if you just want to chat about your codebase.
There is a separate, optional feature called Supermaven Chat which is basically a wrapper around GPT-4o / Claude 3.5 where you can chat about fragments of your code, refactor them etc. This costs money, but you can also supply your own API key for OpenAI/Anthropic models. Or just use the web interface of any LLM if you just want to chat about your codebase.