Last week, OpenAI launched GPT-5, its most powerful model for ChatGPT users and developers. The rollout strategy for GPT-5 created a lot of uproar among ChatGPT users, as OpenAI immediately replaced all existing models, including the popular GPT-4o and o3, with the new GPT-5. Following the feedback, OpenAI made some immediate changes last week.
Today, OpenAI CEO Sam Altman announced some more changes to ChatGPT to satisfy the needs of 700 million weekly active users. First, users can now choose between “Auto,” “Fast,” and “Thinking” modes for GPT-5. When you select Auto in ChatGPT, you’re using a system that can automatically select between Fast or Thinking mode based on your query. Auto will be the default for all users.
To address the complaints about usage limits, OpenAI now offers 3,000 messages/week with GPT-5 Thinking. Once the limit has been reached, ChatGPT will switch to GPT-5 Thinking mini. It is important to note that automatic switching from GPT-5 to GPT-5 Thinking does not count toward the weekly limit. There was some confusion around the context limit for GPT-5 Thinking; OpenAI today clarified that it is 196k tokens.
After last week"s changes, GPT-4o was available in the legacy models list. With today"s update, GPT-4o is now available in the model picker for all paid users by default. Since users were shocked by the decision to suddenly remove the model, OpenAI is now promising that it will give plenty of notice before deprecating it in the future.
For ChatGPT paid users, OpenAI will now have a “Show additional models” toggle in ChatGPT web settings which will list o3, 4.1, and GPT-5 Thinking mini. The GPT-4.5 model will be available exclusively for Pro users.
Finally, OpenAI is also working on an update to GPT-5’s personality to make it feel warmer than its current personality.