GPT-5.2 Optimized for Professional Knowledge Work
Early Release Counters Gemini 3 Pro
Sam Altman: "AGI Will Benefit All of Humanity"
The artificial intelligence (AI) competition between OpenAI and Google is accelerating at an unprecedented pace. OpenAI, the developer behind ChatGPT, has declared an all-out war with Google by unveiling the new GPT-5.2 series just one month after releasing its previous version.
On December 11 (local time), OpenAI announced the launch of the GPT-5.2 series, optimized for professional knowledge work. This comes less than a month after the release of GPT-5.1 on November 13. It is the first time since the launch of ChatGPT in 2022 that OpenAI has released new versions at such a short interval. GPT-5.2 was originally scheduled for release at the end of this month, but the launch was moved up after Google’s Gemini 3 Pro began attracting a large user base. As positive market reviews for Gemini 3 Pro continued, OpenAI CEO Sam Altman issued a "code red" alert within the company and focused on improving ChatGPT’s performance and usability. Google, meanwhile, has enhanced reasoning and multimodal understanding capabilities with Gemini 3 Pro, and is enjoying widespread acclaim for its new image editing tool, Nanobanana Pro.
OpenAI explained that GPT-5.2 is "designed to help people create greater economic value in professional work." The new model demonstrates outstanding performance in spreadsheet creation, presentation structuring, coding, image recognition, understanding long contexts, tool utilization, and executing complex, multi-stage projects. GPT-5.2 features three distinct modes: "Instant Mode" for routine tasks and quick responses, "Thinking Mode" for coding, mathematical solutions, and summarizing lengthy documents, and "Pro Mode" for tackling difficult questions that require extended work time.
OpenAI emphasized that the new version strengthens reasoning and coding abilities, reduces hallucinations, and achieves top-tier results in various AI performance benchmarks. In the GDPval evaluation, which assesses work performance across 44 professional industry roles, GPT-5.2’s Thinking Mode achieved human expert-level or higher performance in 70.9% of categories. Pro Mode scored 74.1%, demonstrating capabilities surpassing those of human experts. GDPval evaluates the ability to generate practical outputs such as sales presentations, accounting spreadsheets, emergency medical schedules, tax filings, manufacturing blueprints, and short videos. In the SWE benchmark, GPT-5.2’s Thinking Mode scored 80%, outperforming Gemini 3 Pro’s 76.2%.
GPT-5.2 also features improved response handling in sensitive conversations, including mental health topics, and applies an age prediction model to enhance content protection for minors under 18, as well as strengthened parental controls. While the subscription price remains unchanged, the cost per token for GPT-5.2 on the API has increased in line with the improved quality. GPT-5.2 will be made available sequentially to paying subscribers starting today, and developers can also access it via API.
This day also marks the 10th anniversary of OpenAI’s founding. CEO Altman stated, "I am almost certain that we will be able to create superintelligence within the next 10 years," and predicted, "People in 2035 will be able to accomplish things that are hard for us to even imagine today." He added, "Our goal is to ensure that artificial general intelligence (AGI) benefits all of humanity. There is still much work to be done, but I am extremely proud of the direction our team has taken."
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
