Assume using ChatGPT 30 days a month, with 100 queries per day.
Mean tokens per query is upper limit (we estimated it using a bunch of queries), in practice this would be lower, especially because it can be bounded by asking ChatGPT to only answer with fixed number of sentences
api_cost_per_token = 2e-6
mean_tokens_per_query = 500
n_queries_per_day = 100
n_days_in_month = 30
n_days_in_month * n_queries_per_day * mean_tokens_per_query * api_cost_per_token3.0