GPT-5 needs option to exclude models from auto switcher :/

Understanding User Experience Challenges with GPT-5’s Auto-Switcher Functionality

Artificial intelligence continues to revolutionize various creative and professional workflows, with models like GPT-5 at the forefront. However, recent user feedback highlights some usability concerns that merit attention for future enhancements.

A Common Pain Point: Unwanted Model Switching During Interactions

Many users, including content creators engaged in creative writing and brainstorming, have reported frustration with GPT-5’s automatic model switching feature. While this feature aims to optimize response quality—particularly through the utilization of specialized “thinking” or “mini-think” models—it can inadvertently hamper efficiency.

For example, during tasks such as generating sensory-rich descriptions of fashion or aesthetics, users often encounter the “thinking” model phase. This process can trigger lengthy response delays or overly verbose outputs—like detailed essays explaining mundane details (e.g., the shape of pants)—which are unnecessary for the task at hand. Such interruptions require users to manually regenerate responses with alternate models, a process that can be time-consuming and disrupt the creative flow.

Impact on Workflow and Resource Management

This need for manual intervention not only interrupts the creative process but also contributes to inefficiencies in thread memory management, especially when multiple regenerations occur. While users have noted that they can set preferred models, the automatic switching mechanism still poses challenges—particularly when users want more granular control over which models are active during specific tasks.

Looking Ahead: The Need for Greater Customization

To enhance user experience, it would be beneficial for platform developers to consider implementing options that allow users to exclude certain models from the auto-switcher or to set strict preferences for model usage during interactions. Providing such controls would give users greater command over their AI interactions, ensuring responses are more aligned with their specific needs—whether that be creative storytelling, technical explanation, or other specialized tasks.

Conclusion

As AI models become integral tools in creative and professional domains, understanding and addressing user-specific preferences is critical. Offering customizable options for model selection within the auto-switcher could significantly improve user satisfaction and workflow efficiency, paving the way for more tailored and effective AI-assisted experiences.