On August 13, OpenAI CEO Altman released an update on the X platform, detailing the latest feature updates of ChatGPT. The core of this update is that GPT-5 has achieved major breakthroughs in multiple key areas.
In terms of response mode, GPT-5 provides three options: "automatic", "fast" and "deep thinking". Altman pointed out that the "auto" mode works for most users and can intelligently match the best way to answer. The new control options provide more convenience for users with specific needs and meet their personalized requirements for answer speed or depth in different scenarios.
In terms of personality shaping, the new version of GPT-5 personality under development will be more friendly. This improvement is based on feedback from many users and aims to avoid the over-enthusiasm problem that GPT-4o had. Altman particularly emphasized that the team is deeply aware of the importance of realizing more personalized AI personality customization functions in the future, and is committed to making the interaction between users and AI more natural and comfortable.
It is reported that GPT-5 adopts an embedded three-in-one integrated architecture, which is composed of the GPT-5-main model for handling conventional problems, the GPT-5-thinking deep thinking model for solving complex tasks, and the routing mechanism responsible for real-time decision-making. When the quota is used up, the mini version will be launched. It can quickly decide which model to activate based on the type of conversation, complexity, tool requirements and the user's clear intention, and independently determine whether to enter deep thinking mode, thereby automatically matching the most appropriate model to answer the question.
According to OpenAI’s public test data, GPT-5’s performance in the fields of mathematics, programming, visual perception and health has greatly surpassed previous generation models.
Currently, GPT-5 is free to use. For higher intelligence levels, you need to pay to upgrade to Plus or Pro users.