Hi - I’m trying to figure out whether this is a problem on the Zapier or ChatGPT side.
When I am calling the API from Zapier it says it is not using half of 8192 tokens and that the output is not being trimmed. However, the output is still being trimmed. See below.
Is there something I’m missing on the Open AI side here? Is there a setting to actually be able to use the 8192 max? I measured the input and output manually using tokenizer and it generally aligns with the numbers that Zapier is showing (i.e. it’s nowhere close to the max).