Understanding the pricing of the ChatGPT API

Greetings, if our institution opts for the Tier 3 plan for the ChatGPT API, allowing students to utilize our API, and the usage either surpasses or falls below the predefined limits of Tier 3, how does it influence the monthly pricing? To clarify, if usage exceeds the Tier 3 limits, will our monthly payment remain fixed at $1000, or will the pricing be adapted according to our actual usage? You can refer to the link provided for details on the tier plans: https://platform.openai.com/docs/guides/rate-limits/usage-tiers?context=tier-three


There’s no such thing as ChatGPT API Tiers.

The API Tiers relate mostly to rate limits; how many requests you can send in a timeframe. https://platform.openai.com/docs/guides/rate-limits

Your tier is determined by how much time you spent on the platform, and the total amount spent. So if you pay $250 each month for 4 months, you’d be in tier 4 possibly in your first month, and tier 5 possibly by the end of your 4th month. if you decide to only spend $10 a month going forward, you’d still be in tier 5, unless or unitl they decide to change it :laughing:

Oh yeah: the tiers have nothing to do with models, as far as I know. Sometimes the best models aren’t available in the free tiers, but the other tier levels have nothing to do with model strength.

Are you kidding me ? Absolutely ridiculous to charge that much , wow what a way to sucker people in to make openai billionaires for what purpose? Shame on tgem

What do you mean? :thinking:

Do you need instant access to a $10_000 spend ceiling and be able to send 2 million tokens per minute, every minute?

The API is pay per use for the actual data that is sent to and from language models, with the costs quite low.

You can engage with AI models for many many interactions with just a minimum prepay credit purchase of $5.

The tier system ensures that your momentary usage and risks from newly funded accounts are mitigated and the usage limits align with the funding history.

The large amounts are just examples that would elevate an account in tier to also being able to regulary spend such large amounts on multi-user product development.

The tiers are based on total previous account payments, not on maintaining usage patterns.

1 Like

Thank you for your replies. If I understand what you have said on the pricing structure for the gpt-4 model for example, the Input costs $0.03 per 1K tokens and the Output costs $0.06 per 1K tokens, if I calculated the total cost using the pay-as-you-use rule for 500 students. Each student averages 30 requests a day, with an input of 100 tokens and receiving an output of 1000 tokens. The total cost is determined as follows:

Total Cost=500×((1/1000×100×0.03)+(1/1000×1000×0.06))×30

After calculation, the total cost amounts to $945. Could you kindly provide me more information on the tiers and how they may impact costs in this case.

Tiers don’t impact cost!

Only your rate and spend limits!

If you expect to spend this much but have a low tier account, consider requesting an exception at the bottom of https://platform.openai.com/account/limits

It might take a while for OpenAI to get back to you, but they tend to eventually.

1 Like

@Diet, @_j many thanks for your explanation!

1 Like