ChatGPT now offers 3 new GPT-5 settings and legacy models
What's the story
OpenAI launch of GPT-5 was billed as a major leap in artificial intelligence, promising to streamline the ChatGPT user experience with a new universal model router. The idea was that users would no longer need to sift through multiple model choices. Instead, GPT-5 would automatically select the best AI engine for each query. But reality quickly diverged from this vision. Now, within days of GPT-5 launch, OpenAI has reverted to the old model picker after widespread user dissatisfaction.
User options
Paid subscribers have regained access to legacy models
Instead of eliminating choices, OpenAI reinstated the model picker in ChatGPT. Users can now select among 'Auto', 'Fast', and 'Thinking' modes, each representing different priorities for speed or depth of AI response. The 'Auto' mode uses the new GPT-5 router, while the other options let users control how their queries are processed. Paid subscribers have also regained access to legacy models, such as GPT-4o, GPT-4.1, and even GPT-3, which had been briefly deprecated.
Future improvements
GPT-5's personality update
OpenAI CEO Sam Altman revealed that the company is working on an update to GPT-5's personality. He said it should be "warmer than the current personality but not as annoying (to most users) as GPT-4o." The CEO also stressed the need for more per-user customization of model personality in future iterations of their AI models.
Future plans
Altman addresses user concerns
In light of the backlash, Altman has promised that the company will give users plenty of advance notice if it ever decides to remove GPT-4o. He also acknowledged the challenges in routing prompts to the right AI model, which involves aligning an AI model with a user's preferences and their specific question. The removal of GPT-4o and other AI models from ChatGPT drew criticism from users who had grown fond of the responses and personalities of these models.