OpenAI is testing new ChatGPT Intelligence settings

An experimental configuration screenshot of the ChatGPT interface revealed a new Intelligence settings that will allow users to customize default model versions and “thinking effort” levels.

A screenshot shared by Testing Catalog shows the entire page called “Intelligence.” It features a model selector showing “Latest – 5.2 v” as an option alongside standard settings like “Auto.” A separate “Thinking effort” dropdown is set to “Extended,” with a note explaining: “More effort can give more thorough answers, but replies can take longer.”

Additional toggles include “Thinks longer for better answers” and a “Pro” upgrade prompt for enhanced capabilities. Users can even “Set as default” for persistent preferences.

These settings represent a significant step toward user empowerment. Currently, ChatGPT relies on automated model selection (e.g., GPT-4o or o1 variants), but explicit controls could let users balance speed, accuracy, and depth. “Extended” thinking effort aligns with OpenAI’s chain-of-thought techniques seen in reasoning models, promising richer responses for complex tasks like research or coding.

Some early users noted similar features appearing in A/B tests weeks earlier. If rolled out widely, this could make ChatGPT more versatile, appealing to casual users who need quick replies and power users demanding maximum intelligence. However, we’ll have to wait for an actual rollout.