Feature Request: Add Temperature / Top_p Controls for Custom GPTs

[flacora-projetos]

Hi OpenAI Team,

I’d like to formally suggest the implementation of a temperature (and optionally top_p) setting in the ChatGPT interface for Custom GPTs.

As the adoption of GPTs grows across professional and enterprise use cases, many of us are building Custom GPTs to automate decision-making, reporting, or other critical workflows. These contexts require high reliability, low randomness, and deterministic outputs.

Currently, without control over generation parameters like temperature or top_p, we are unable to guarantee consistent behavior. This limits the applicability of GPTs in business scenarios where factual accuracy and output stability are essential. When errors or hallucinations occur due to high randomness, the core value of automation — trust and efficiency — is undermined.

Feature request:

• Add a temperature slider (range 0.0 to 1.0) in the Custom GPT setup screen.
• Optionally include a top_p setting for additional tuning.
• Persist these values per GPT (not per session).

These simple additions would empower builders to tailor their agents to real-world applications more effectively, while avoiding unintended variation in behavior.

Thank you for considering this improvement. It would be a major step toward making Custom GPTs more production-ready.

Best regards,
Flávio Corá - Brazil