OpenAI has presented its new artificial intelligence (AI) model GPT-5.2the most advanced yet for professional tasks with Instant, Thinking and Pro versions for ChatGPT introducing the following Reasoning, coding and security improvementsby ensuring fewer hallucinations and better protection of mental health.
The company has more than 800 million weekly users in their chatbot, which accounts for around 70 percent of the assistant’s usage, as well as more than Four million developers are building on the OpenAI platform.
In this spirit, to continue offering technology that advances everything professionals, teams and companies want to reliably achieve with AI, OpenAI has introduced its new flagship model GPT-5.2, less than a month after its predecessor GPT-5.1.
The technology company describes this model in both the ChatGPT assistant and the API as “the best yet” for everyday professional use. Thus, this update of the model includes improvements in several aspects that “set a new standard” in reference assessment.
As detailed in a statement, the GPT-5.2 family has been integrated into ChatGPT with versions GPT-5.2 Instant, GPT-5.2 Thinking and GPT-5.2 Pro significant improvements for work and learning. In the case of the instant version it offers faster and better answers to information search questionsStep-by-step instructions, technical texts and translations.
Likewise, GPT-5.2 Instant is also more effective in this regard Intended use to support learning and skills developmentas well as to provide clearer work and professional guidance.
On the other hand, with GPT-5.2 Thinking, OpenAI has highlighted the improvements introduced that make it the “best model for professionals yet”. This is because its ability to solve more complex problems with more precision has evolved.
OpenAI, for example, pointed this out Improvements to spreadsheet organizationas well as the Progress in creating presentations. All this is supported by its results in relevant tests such as GDPval, the “benchmark” that, in addition to the tasks mentioned above, evaluates 44 real professional tasks, including document writing, in which it achieves “expert level” points, establishing itself as the first model to achieve this.
Specifically, across the 44 tasks in this test, GPT-5.2 Thinking outperformed or equaled the best professionals in the industry 70.9 percent of the time, while GPT-5 Thinking only achieved this 38.8 percent of the time.
Finally, the GPT-5.2 Pro version appears to be “the smartest and most reliable model for difficult questions.” This option has a longer response time but offers higher quality results. As a result, in preliminary tests it showed higher accuracy, fewer fatal errors, and superior performance in complex areas such as programming.
Performance and safety
In addition, the new model also excels in coding tasks, where GPT-5.2 Thinking reaches 55.6 percent in SWE-bench Pro, in scientific reasoning tasks, improving results according to GPQA Diamond, and in mathematics tasks, where it reaches 40.3 percent in FrontierMath.
Another improvement to consider is that GPT-5.2 Thinking produces fewer changes than GPT-5.1 Thinking, OpenAI in particular has highlighted this Wrong answers were 38 percent less common. This allows users to use ChatGPT with greater reliability when researching, writing, analyzing or making decisions in daily work.
In this sense, security advances in the use of the chatbot have also been expanded, with a particular focus on mental health. This is because GPT-5.2 is now available More accurately detects possible signs of a user’s discomfort and take care of your answers to offer help. This security improvement was implemented after responsibility was denied for the suicide death of a teenager who used ChatGPT as a confidant.
On the other hand, OpenAI has explained in detail that the new GPT-5.2 is represented in the API with justification the best option for developers when creating agents.
This is because it offers the following Significant improvements in general intelligence and comprehension of long textsthe autonomous use of tools by the agent and vision, as it reduces the margin of error in graphic reasoning and can more accurately interpret diagrams or the way elements are placed in an image, as the company has stated.
In this way, the Sam Altman-led technology company ensures that GPT-5.2 outperforms any of its predecessor models in fully executing complex real-world tasks.
With this in mind, OpenAI has announced that it has begun deploying GPT-5.2 (Instant, Thinking and Pro) to users subscribed to the Plus, Pro, Business and Enterprise plans. Implementation will also take place gradually.
OpenAI unveiled this new model a week after the company was declared “code red” under pressure from its competitors, particularly Google. At the time, Altman stated that he would restructure his efforts to improve the quality of ChatGPT, with improvements to user customization features and increased speed, before focusing on other products. He also announced that he would launch a new reasoning model that would outperform Google’s Gemini 3 model.
About The Author
