ChatGPT moves quickly to end support for most models
New UI hides model selection behind tiny arrow, defaults all chats to 'Latest' with auto-switch.
OpenAI has implemented a major interface overhaul for ChatGPT on web and iOS that makes selecting specific AI models like GPT-4 significantly more difficult. The new design hides the model picker behind a tiny, non-intuitive downward arrow, a departure from the previously clear selection menu. Crucially, the system no longer remembers a user's model choice, forcing every new chat to default to 'Latest + Auto-switch to Thinking.' This change effectively pushes all users toward OpenAI's recommended, constantly updated model.
The update strategically marginalizes older 'Legacy' models. These options are now buried within a confusing, multi-step modal and dropdown system, accompanied by prominent warnings urging users to revert to 'Latest.' Industry observers and software engineers note this is a classic product tactic: make a feature hard to find and use, then use the subsequent drop in usage data to justify its removal. The move aligns with CEO Sam Altman's previously stated vision for a single, 'evergreen' ChatGPT experience, suggesting a future where paid tiers (Plus and Pro) may only offer one or two model choices.
- Model picker is now hidden behind a tiny downward arrow, a major reduction in discoverability.
- User model selections no longer persist; every chat defaults to 'Latest + Auto-switch to Thinking.'
- Legacy models are labeled with warnings and buried in a complex, multi-step interface.
Why It Matters
This signals a shift away from user choice, potentially locking professionals into a single, opaque 'Latest' model for core workflows.