Assume I use the gpt-5 model on medium reasoning (med verbosity, summary auto),
I use web_search (non preview), and I see 60 sources being used for that request (10k input tokens, 1k output tokens)
am I charged ONLY for the input output tokens, meaning that the 60 sources are baked into the input (i.e. the 10k tokens, assuming my prompt is just 500 tokens) + the output tokens?
Specifically, what I want to understand is how much am I being charged per source. if that $10/1k calls the same as $10/1k sources?
I don’t get it.
more specifically, how much would I be charged here?
gpt-5 model
