• Instructions to Ask a Question

    For any assistance, please click the "Ask a Question" button and select the Pabbly product for which you require support.

    We offer seven comprehensive applications designed to help you efficiently manage and grow your business:

    Our support team endeavors to respond within 24 business hours (Monday to Friday, 10:00 AM to 6:00 PM IST). We appreciate your understanding and patience.

    🚀 Exclusive Lifetime Offers 🚀

    We invite you to take advantage of our special one-time payment plans, providing lifetime access to select applications:

    • 🔥 Pabbly Connect — Lifetime Access for $249View Offer
    • 🔥 Pabbly Subscription Billing — Lifetime Access for $249View Offer
    • 🔥 Pabbly Chatflow — Lifetime Access for $249View Offer

    Make a one-time investment and enjoy the advantages of robust business management tools for years to come.

Pabbly ChatGPT Structured Output: GPT-4.1 Max Tokens Setting Ignored in Pabbly?

adampj

Member
I’m running into another parameter issue, this time with gpt-4.1-2025-04-14 (or GPT-4.1) when using the ChatGPT (Structured AI Output) action in Pabbly. I set the Max Tokens field to 15000 (screenshots attached), but the responses keep getting capped at 2048 tokens, which seems to be the default. The debug output confirms that Usage Completion Tokens never exceeds 2048, even though the model supports much higher limits.

This suggests that the Pabbly integration isn’t passing the max_tokens parameter correctly (or at all) to the API, and is instead defaulting to 2048. It looks like the same underlying issue as with GPT-5 — the field labeled “Max Tokens” is either ignored or mapped incorrectly.

Can Pabbly confirm whether max_tokens is supported in the current integration? And if not, is there a workaround to request completions above 2048 tokens (e.g., custom API call, mapping override, or update to the connector)?

Screenshots:

Screenshot 2025-09-06 at 17.07.38.pngScreenshot 2025-09-06 at 17.07.20.png
 
Last edited:

ArshilAhmad

Well-known member
Staff member
This appears to be a different workflow related to your other issue. It is using the GPT-5 AI model. Please share the workflow related to this issue.
 
Top