Why does it seem that the cost of input is higher than the cost of output when looking at the Usage list? Is it reversed?
Hey @raymondreamer!
TL;DR:
Input tokens = everything you send to the AI.
Output tokens = what the AI writes back.
You pay $2 per million input tokens and $10 per million output tokens. In your case, most of the cost came from how much you sent in, not what the AI replied with.
More detailed explanation (expand)
Basically, the number of input tokens is data sent into the model, which would include the model’s system prompt, tools, the contents of files it views and edits, your messages, and more. The number of output tokens is how many tokens the model generates itself, which is usually much lower than the number of input tokens for agentic tools like v0. Claude models, which v0 uses, also limit the output tokens to 64k.
The pricing card in your second image is saying, “for every 1 million input tokens, you get charged $2, and for every 1 million output tokens, you get charged $10”. So in your second image, it means you used about 0.62 million input tokens (which costs $1.24) and about 13,000 output tokens (which costs $0.13).
2 Likes
I understand what you’re saying, but in reality, I don’t have that many inputs in this conversation; rather, I have more outputs. So, I suspect that the usage in the tooltips is reversed.
1 Like
You probably have a lot more input than you think! Input isn’t just the messages you send. Plus, it compounds.
You are quite possibly right though, so hopefully the team will take a look at this just to make sure 
1 Like