OpenAI CEO Sam Altman announced new updates for GPT-5 on August 13, introducing three user-selectable response modes: Auto, Fast, and Thinking. The default mode remains Auto, but Altman believes the extra flexibility will benefit some users. The “Thinking” mode now has a weekly limit of 3,000 messages, with additional access provided through GPT-5 Thinking mini. A major enhancement includes a 196k-token context limit, enabling GPT-5 to handle much longer conversations and documents. Altman mentioned that usage patterns might lead to adjustments in this limit or message caps in the future.
GPT-4o has also made a return as the default model for all paying users. Altman assured users that if OpenAI ever phases it out, sufficient advance notice will be given. A new “Show additional models” toggle in ChatGPT’s settings allows users to access models like o3, 4.1, and GPT-5 Thinking mini. GPT-4.5 remains available only to Pro users due to its high GPU demands.
OpenAI is also refining GPT-5’s personality to make it warmer yet less overwhelming than GPT-4o. The company emphasizes the need for greater personalization in future versions. GPT-5 is touted as their most advanced model yet, especially in coding and complex reasoning.
