How can I fine tune a gpt model

Assume I use OpenAI Platform by providing my own data (there is even an example on doing that in a subsequent paragraph of the same guide) : will this action trigger an update of 175B parameters? Where is the new model stored? Will this become a new model available for others to work on? (i.e. doing inferences) ? How correct is the belief that open source LLM models are cheaper to fine tune ?