Openai announced the launch of its new GPT-5 artificial intelligence model. It has become more precise, less likely to make mistakes, answers faster and safer. GPT-5 is already available for users with Plus, Pro, Team tariffs and even free accounts. Enterprise and Edu owners will gain access in a week.
The company introduced three versions: GPT-5, faster and compact GPT-5 mini, as well as an improved GPT-5 Pro with expanded reasoning opportunities. Free users will be able to use the GPT-5 limitedly, after which they will switch to the GPT-5 mini. Plus subscribers have more limit, and Pro has no restrictions.
The main innovation-GPT-5 is initially a “model of reasoning”, that is, he knows how to step by step to analyze complex tasks and find a solution. Compared to the GPT-4O, the number of actual errors decreased by 45%, and with earlier versions-up to 80%. The model has become more honest: it will not come up with the performance of the task, confidently respond with doubts or adapt to the user’s opinion.
The SAFE Complets also appeared – now AI is trying to give a useful answer within the framework of safe rules, and not just agree or refuse your opinion.
In addition, Chatgpt received four new “personalities”: cynic, robot, listener and nerd. While they are available only in text chat, the voice will appear later.
