Recently, OpenAI has made significant upgrades to ChatGPT, optimizing model selection, rate limits, and user experience based on user feedback. According to the latest information compiled by AIbase, this update not only restored GPT-4o as the default model but also introduced multiple mode options for GPT-5, while optimizing AI personality to make it warmer and more user-friendly. This update provides users with a more flexible and efficient AI interaction experience.
GPT-5 Modes Offer Flexible Switching to Meet Diverse Needs
The new version of ChatGPT allows users to freely choose between three modes of GPT-5: "Auto," "Fast," and "Think." The "Auto" mode intelligently allocates resources based on task complexity, the "Fast" mode is suitable for scenarios requiring immediate responses, and the "Think" mode is designed for complex tasks, emphasizing deep reasoning capabilities. AIbase learned that the context window of the GPT-5 Thinking mode can reach up to 196K tokens, enabling it to handle longer and more complex conversations. However, the rate limit for this mode is 3000 messages per week, after which it will switch to the lightweight GPT-5 Thinking mini mode to ensure service stability.
GPT-4o Returns, Model Selector Fully Optimized
In response to user feedback about the initial experience of GPT-5, OpenAI quickly made adjustments, restoring GPT-4o as the default model for all paying users. Paying users can also access various models, including o3, 4.1, and GPT-5 Thinking mini, through the "Show Other Models" option in the ChatGPT web settings. Additionally, GPT-4.5, as a high-performance option, is only available to Pro users due to its high demand for GPU resources. This change allows users to switch models flexibly according to their needs, greatly enhancing the freedom of the user experience.
AI Personality Upgrade, More Warm Interaction
This update also optimized the interactive personality of GPT-5, aiming to make the AI more warm and user-friendly while maintaining professionalism. Compared to the previous GPT-4o, the optimized GPT-5 personality is described as "friendlier and less likely to annoy users." This adjustment responds to some users' criticisms that the early GPT-5 was too "stiff" or "corporate," aiming to provide a more natural and human-like conversation experience.
Visual and Interactive Improvements, Gradient Interface Removed