ChatGPT moves quickly to end support for most models

ChatGPT Discontinues Support for Most Models in Favor of a Single Evergreen Option

In a recent update, ChatGPT has introduced a new user interface for selecting AI models, both on its web platform and iOS app. This change appears to align with a strategic shift towards streamlining options, reflecting what industry observers understand as a move by OpenAI’s leadership—particularly Sam Altman—toward promoting a single, consistently-updated model as the primary choice for most users.

The Design Changes: An Obvious Direction

As a software engineer analyzing these updates, it’s clear that the changes serve to phase out older or less-used models by making them less visible and harder to access. Traditionally, users could easily select from a range of models; now, the interface is subtly redesigned to push users toward the “Latest” model, which benefits OpenAI by focusing usage data on one main model.

Here’s what’s happening behind the scenes:

• Model Selection is Hidden: The new model picker is concealed behind a small downward-pointing arrow—often overlooked by users. Previously, this menu displayed the current model (e.g., “ChatGPT Auto”), but now accessing other options requires additional steps.

• User Preferences Are Not Persisted: Earlier, the model choice would carry over to subsequent conversations, providing consistency. The new interface resets to a default setting—“Latest” with auto-switching to “Thinking”—regardless of prior selections.

• Complex and Reduced Usability: What was once a straightforward list of models now involves multiple interface elements—dropdowns, modals, and separate lists—fragmenting the experience and making it less intuitive.

• Legacy Labels and Prompts: Other models are now marked as “Legacy” with a prominent prompt encouraging users to revert to the “Latest” model, subtly discouraging ongoing use of the older options.

Implications for User Data and Model Popularity Metrics

These interface modifications are likely aimed at influencing user behavior—intentionally or not—by decreasing the visibility and accessibility of less recent models. Over a period of roughly a month, data collected by OpenAI may show a sharp decline in the usage of older or alternative models, not necessarily due to user preference but due to reduced discoverability.

As a result, the company may report a minimal percentage of users engaging with different models, effectively consolidating usage statistics around a single, “best-supported” model. Furthermore, for paying customers on Plus and Pro plans, OpenAI appears poised to further streamline choices down to only two options, minimizing complexity at the expense of flexibility.

Conclusion

This strategic interface redesign exemplifies a broader trend in platform simplification—prioritizing a single dominant model and reducing the visibility of alternative options. While this approach can help ensure widespread adoption of a superior, regularly updated model, it also raises questions about user choice, transparency, and the genuine utility of maintaining multiple models.

As users and developers, understanding these design choices is crucial for navigating the evolving AI landscape and advocating for interfaces that balance simplicity with meaningful flexibility.