I am currently using gpt-3.5-turbo and based on my calculations, I will be paying over $100 per day for the API fees once my new project goes live. This feels daunting and unaffordable to me.
Here is how I calculated the cost: I need to store a prompt text in the system and I have estimated that the token count for this text will be around 3000. Based on the gpt-3.5-turbo pricing of $0.002 / 1K tokens, each access will consume $0.006. Assuming one access every 5 seconds, the daily access count will be 246060/5=17280. Multiplying the cost per access, $0.006, by the daily access count, we get $103.
So this is my calculation. Is there any way to reduce the cost while still passing enough information to the system? I feel like the popularity of GPT-3 will depend on the cost of usage and whether we as developers can find a way to balance our costs and profits.