Max tokens Response length limit

Hello,

You are partly correct.

The Max tokens setting limits the maximum number of output tokens ai can generate. It does not limit the input tokens or the combined total.

Units means:

Input tokens + output tokens = total units

Your pricing rule calculates credits as follows:

(input tokens ÷ 1000 × 10) + (output tokens ÷ 1000 × 20)

Based on 36583 units and a charge of 392 credits, the approximate breakdown would be:

Around 34000 input tokens
Around 2500–2600 output tokens

That would be approximately 340 input credits plus 52 output credits. Therefore, the output was well below the 10000-token limit—the 36583 figure does not mean the response itself contained that many tokens.

The high input usage could come from the form instructions (your prompt), submitted content, system prompt, or additional context such as web search.

You would need to inspect the corresponding usage/conversation log to identify the exact contributor under Usage module.

Just a note: I see that your max token is 10000. It is quite high for most forms. For standard responses, I recommend starting around 1000–3000 and increasing it only if responses are being cut off.