Confused about pricing on the 3.5 Turbo model

API
1

I have been using the gpt-3.5-turbo model most recently and as I recall the pricing was 0.0015/1k prompt tokens and 0.002/1k completion tokens. Today Open AI announced pricing cuts and for 3.5-turbo, the prices are now 0.003 for prompt and 0.004 for completion, but that is MORE, not a price cut, so I must be misunderstanding something. Then I go on the pricing page and the values are 0.001 for prompt and 0.002 for completion which has a decrease for prompt only.

So what is the actual price and what am I missing with the announced pricing cut cost which is higher?

2

You will note in pricing that the price is only highlighting the new model names:

Model Input Output
gpt-3.5-turbo-1106 $0.0010 / 1K tokens $0.0020 / 1K tokens
gpt-3.5-turbo-instruct $0.0015 / 1K tokens $0.0020 / 1K tokens

The price also likely has a direct correlation to the amount of computation given to your response generation. So see what these models do before flipping a switch.

3

That’s fine-tuned prices, I think?

4

Fine tune:

Model Training Input usage Output usage
gpt-3.5-turbo $0.0080 / 1K tokens $0.0030 / 1K tokens $0.0060 / 1K token

The only thing he could be referring to is the unchanged price of gpt-3.5-turbo-16k if invoking that model.

Should you just switch?

Testing: GPT-4 works on this old preset:

image
image903Ă—547 12.8 KB

gpt-3.5-turbo-0301 worked and works still despite an IQ hit if you find the json

image
image902Ă—259 9.32 KB

gpt-3.5-turbo-16k-0613: horrifically broken two months ago:

image
image897Ă—285 8.27 KB

gpt-3.5-turbo-1106 - remains hosed

image
image899Ă—293 8.38 KB

As you can see in the prompt, this was typed up for the forum as a “here’s how you can write a classifier in two minutes knowing what I know about what works the first time”.

1 Like