ChatGPT’s model picker is back, and it’s complicated

OpenAI recently unveiled GPT-5, aiming to enhance the ChatGPT experience with a streamlined model approach. The expectation was for GPT-5 to serve as a versatile AI solution, automatically selecting the best response method for users, thereby eliminating the need for a complicated model picker. However, early feedback suggests that the rollout has not met these ambitious goals. In a recent update, OpenAI CEO Sam Altman announced the introduction of three distinct settings: 'Auto', 'Fast', and 'Thinking'. While 'Auto' is designed to function as the intelligent router initially promised, users can also bypass this feature to choose models with varying response speeds. This added flexibility may appeal to users looking for specific interaction styles, yet it also complicates the user experience. With the launch of GPT-5, the messaging limit has increased to 3,000 messages per week for the 'Thinking' setting, offering enhanced capacity for dedicated users. Additionally, Altman confirmed that paid subscribers can once again access previous legacy models like GPT-4o, which had been removed shortly before GPT-5's introduction. This change reflects user demand for familiar AI personalities and responses. In his communications, Altman acknowledged the necessity for improved customization in AI personalities, promising a warmer tone in future updates while avoiding the pitfalls of previous models. Despite these efforts, the complexity of the model picker has persisted, leaving many users dissatisfied. The transition to GPT-5 has not been seamless. Users expressed frustration over the sudden removal of preferred models, and the initial malfunction of the model router raised concerns about performance. In response, Altman addressed these issues during an AMA session on Reddit, highlighting the challenges in aligning AI models to user preferences rapidly. As the landscape of AI interactions evolves, OpenAI recognizes the unique attachments users form with specific models. This phenomenon has been illustrated recently, as a group in San Francisco held a symbolic funeral for Anthropic's Claude 3.5 Sonnet after its discontinuation. OpenAI is now tasked with further refining its AI offerings to better align with user needs and preferences.