Altman: Introducing the GPT-5 model switch, premium users can choose between 4.0 and 4.5, which is only available to Pro users.

OpenAI CEO Sam Altman: Now you can choose Auto, Fast, or Thinking for GPT-5. Most users will choose the Auto model. GPT-5 Thinking has a weekly quota of 3,000 tokens, which can be supplemented with additional capacity from the Thinking mini model. The context limit for GPT-5 Thinking is 196,000 tokens. We may need to update the rate limit over time based on usage. 4o has now been default re-added to the model selection for all paid users. If we plan to deactivate this model in the future, we will notify users in advance. Paid users can also find the "Show more models" switch in the web settings, which allows them to add models like o3, 4.1, and GPT-5 Thinking mini. 4.5 is only open to Pro users and requires a large amount of GPU resources. We are updating the personality of GPT-5 to be warmer than its current personality, but not as annoying as 4o to most users.