OpenAI introduced GPT-5.2, its most advanced artificial intelligence model to datein a launch that also directly impacts the millions of users who use ChatGPT in Spain to study, work or automate daily tasks. The company ensures that this new generation gives a leap in accuracy, reliability and reasoning abilityboth in the wizard and in its developer API.
This announcement comes amid an escalation of competition in the sector, just one week after OpenAI declared the The “red code” under pressure from competitors like Google. Sam Altman has already announced that he will prioritize the improvement of the assistant, a move particularly relevant in countries like Spain, where the professional use of ChatGPT continues to grow and companies are increasingly integrating these models into their daily activity.
Three versions: Instant, Thinking and Pro
GPT-5.2 is coming three modalities designed for different types of use in ChatGPT.
It is the fastest version and is geared towards everyday life: information searches, step-by-step guides, translations, technical writing and study assistance. OpenAI explains that it generates clearer and more agile responses, designed to increase productivity and facilitate learning.
This is the model that is evolving the most and the one that OpenAI considers “the best for professionals”. Improve complex problem solving, spreadsheet organization, presentation creation, and advanced document writing. In the GDPval test – which evaluates 44 real-world work tasks – he outperforms or equals top experts 70.9% of the time. Its predecessor, GPT-5.1 Thinking, only achieved 38.8%.
It is the most accurate alternative for difficult questions and highly technical areas. It offers higher quality responses, but with longer wait times. In prelims, you achieve greater accuracy, make fewer major mistakes and above all excel in programming.
A leap into reasoning, coding and mathematics
OpenAI has also significantly improved the model’s technical testing capabilities. In coding, GPT-5.2 Thinking reaches 55.6% in SWE-bench Pro. In scientific reasoning, it improves over GPT-5.1 according to GPQA Diamond. In mathematics, he progressed up to 40.3% in FrontierMath.
Another key point is error reduction: GPT-5.2 Thinking generates 38% fewer incorrect answers than its previous version. According to the company, this results in more reliable use for research, writing, analysis and decision-making in daily work.
In the API, GPT-5.2 with reasoning allows you to create more intelligent and autonomous agents. Improved understanding of long texts, independent use of tools and interpretation of graphics and diagrams. It is also more precise when analyzing images and arranging its elements. The company claims that is your best model for performing complex real-world tasks.
More security and special attention to mental health
The model incorporates security improvements, with a special emphasis on detecting signs of emotional distress. GPT-5.2 more accurately identifies sensitive situations and adapts its responses to provide more appropriate help.
The update comes after OpenAI will deny responsibility in the case of a teenager who committed suicide after using ChatGPT as a confidant.
Availability
GPT-5.2 in its Instant, Thinking and Pro versions is already gradually being deployed to users of the Plus, Pro, Business and Enterprise plans. The launch follows Altman’s announced strategy to strengthen the quality of the assistant and improve personalization and speed.
The company claims that this model outperforms all those launched so far and that its reasoning performance surpasses that of Google’s Gemini 3 model, its main rival at the moment.
About The Author
