ChatGPT Model Picker Returns Amid GPT-5 Rollout Reactions

OpenAI’s recent launch of GPT-5 was intended to simplify the ChatGPT experience, but the return of the model picker and additional options have complicated matters.

OpenAI’s release of GPT-5 was expected to streamline the ChatGPT experience by introducing a model router that would automatically determine the best AI model for users based on their prompts. The goal was to eliminate the model picker, a long and complex menu of options that had frustrated many users, including OpenAI CEO Sam Altman, who publicly expressed his dissatisfaction with it.

However, despite the high expectations for GPT-5 to simplify the process, it appears that the new system has not met all user expectations. Instead of streamlining the experience, OpenAI has once again introduced a model picker, albeit with additional features. Users now have the option to select between several settings in GPT-5, namely Auto, Fast, and Thinking, which offer different types of responses based on the user’s needs.

The New Model Picker and Its Settings

While Auto mode is designed to function similarly to the original idea of a model router, allowing GPT-5 to decide how best to respond, the other two modes, Fast and Thinking, offer users more control over how they interact with the AI.

• Auto Mode: This mode still utilizes the model router, but it’s now just one option among several. GPT-5 uses this setting to choose the most appropriate model based on the prompt provided.
  
• Fast Mode: This setting prioritizes speed, selecting an AI model that provides quicker responses, often at the cost of detailed or thoughtful answers.
  
• Thinking Mode: As the name suggests, this setting allows GPT-5 to take more time in generating a response, focusing on delivering more thorough, in-depth answers, even if it means a slower response time.
  

The inclusion of these additional options suggests that OpenAI is giving users more control over how they interact with GPT-5, acknowledging that some may prefer faster responses, while others want more detailed and thoughtful replies. The company is responding to feedback from users who value the ability to choose between these different modes.

The Return of Legacy Models

Alongside the introduction of the new settings, OpenAI has reintroduced several legacy models that had previously been deprecated. These models include GPT-4o, GPT-4.1, and o3, which had been removed in earlier versions of ChatGPT. GPT-4o, in particular, has become a fan favorite for its unique personality and response style, and many users were upset when it was removed.

As part of the update, GPT-4o has been made available by default in the model picker for paid users, with the option to re-enable other legacy models through ChatGPT’s settings. This move has been a response to user feedback, as many had become emotionally attached to these older models and their distinctive conversational styles.

In addition to reintroducing these models, Sam Altman and other OpenAI executives have acknowledged the emotional attachment users have developed to certain AI personalities. Altman has promised that in the future, users will be given plenty of notice before any models are deprecated, in an effort to avoid backlash and user dissatisfaction.

GPT-5’s Challenges: Model Router and User Preferences

Despite the excitement around GPT-5’s new features, the rollout has not been without its challenges. The initial launch of GPT-5 and its model router didn’t live up to expectations, with some users reporting that the system wasn’t as effective as anticipated. 

The router was supposed to automatically assign the best model for a given prompt, but many users felt that GPT-5 wasn’t performing as well as earlier models, prompting OpenAI to address the issue in a public AMA session.

One of the main challenges with GPT-5’s model router is routing the prompt to the right AI model, which involves a complex decision-making process that must consider not only the user’s question but also their preferences. 

For instance, if a user prefers a fast response rather than a more detailed answer, the router needs to account for that. Additionally, the router must decide quickly, selecting between fast and slow models to ensure that the AI can respond within a reasonable time frame.

While OpenAI has improved its model router, many users still prefer to be able to directly choose which model they want to interact with based on their current needs. This preference for user control reflects a broader trend in AI development, where consumers want more personalization and customization in their interactions with AI systems.

User Preferences and AI Personalities

One key takeaway from the launch of GPT-5 and its associated features is how much AI personalities matter to users. 

Over time, users have developed attachments to specific AI models and their unique ways of responding. For instance, some users appreciate concise, fact-driven responses of particular models, while others prefer the more conversational, engaging tone of others.

OpenAI has acknowledged this growing attachment to AI personalities and has hinted at plans for more personalization in future versions of GPT models. 

Altman has stated that OpenAI will focus on creating models more adaptable to user preferences, including a more “warm” personality for GPT-5. However, striking the right balance between warmth and professionalism has proven difficult, as users’ preferences vary widely.

The ability to personalize AI interactions is becoming an increasingly important feature for many users. This is especially relevant when users use AI for complex, long-term tasks, such as software development or content creation, where they may develop a rapport with the AI.

The Future of AI Customization and User Control

As OpenAI continues to refine GPT-5, personalization and user control will likely become even more central to the company’s development process. Giving users more freedom to customize their experience could help eliminate some of the frustration of navigating multiple model options. 

OpenAI’s long-term goal is to create a world where users don’t need to adjust settings or manually choose models, and the AI can seamlessly provide the best response automatically.

This vision for AI customization also extends to the broader AI ecosystem, where companies like Anthropic and Google are exploring ways to give users more control over their AI interactions. As AI technology evolves, the ability to personalize interactions will likely become a defining feature of the industry.

OpenAI’s Ongoing Struggles and User Demands

The launch of GPT-5 has faced some challenges, particularly regarding the model router’s ability to meet user expectations. The return of the model picker and the introduction of new settings demonstrate OpenAI’s effort to give users more control over their interactions with AI, even as it continues to refine its approach.

While GPT-5 has not fully realized OpenAI’s vision of a unified, one-size-fits-all AI model, the introduction of legacy models and new customization options has been a step in the right direction. 

Moving forward, OpenAI will need to continue addressing the complexity of user preferences and finding ways to balance AI performance with personalization. The company’s commitment to improving its models and listening to user feedback will likely shape the future of ChatGPT and its role in the AI landscape.