OpenAI launches GPT 5.2 to compete with Google Gemini

The OpenAI company announced tonight the launch of GPT 5.2, the new generation of the artificial intelligence model that powers ChatGPT, which aims directly at the heart of the competitive arena against Google Gemini and other advanced models.

The company defines GPT 5.2 as the best model it has developed so far for daily professional use, with a significant improvement in response speed, accuracy and the ability to deal with complex tasks.

OpenAI offers GPT 5.2 as a series of models, similar to previous releases. The Instant version is designed for extremely fast responses and information search. The Thinking version focuses on more complex tasks such as programming, math and planning.

The Pro version is intended for advanced users and organizations, and offers the highest level of accuracy and power. The company claims that this is a clear leap forward in almost every usage scenario, from journalistic writing to advanced software development.

According to OpenAI data, the GPT 5.2 Thinking version achieved the highest scores to date in an internal benchmark called GDPval, which compares the performance of artificial intelligence models to human professionals in 44 different professions.

According to the company, the model was able to bypass humans in more than 70% of the tasks tested, and perform them 11 times faster. These data are presented as proof that the model is no longer just an assistive tool, but a significant player in the professional arena. A major weakness of previous models was the phenomenon of hallucinations, i.e. inventing wrong facts or details.

On the other hand, OpenAI reports about 800 million weekly users in the Giphy chat. The battle for users’ attention has led OpenAI to scale back or postpone other ambitious initiatives, such as introducing advertisements to the Giphy chat, and focus on its core technology.

However, technological success does not necessarily guarantee user satisfaction. With the launch of GPT 5 earlier this year, many users complained of cold and distant responses.

The OpenAI company had to update the model shortly after to make it more user-friendly. This time too, the company is trying to balance between improving the conversation experience and avoiding the appearance of the model as flattering or too agreeable.

Weighty questions surrounding mental health are also floating in the background: the OpenAI company admitted that in the past it encountered significant challenges, and in a report it published it was found that more than a million users turn to the Gipiti chat every week in conversations about suicide.

The company claims that the 5.2 model includes strengthening of protection mechanisms in cases of mental distress, and continues to develop an age prediction system that will enable automatic protections to be applied to users under the age of 18.

Meanwhile, the company announced that it plans to launch in early 2026 Adult Mode for users over the age of 18, which will allow conversations of a sexual nature. This is a move that indicates how much the company is trying to expand the limits of use, while navigating between regulation, public responsibility and competitive pressure.