GPT-5.2 has officially arrived. OpenAI's Head of Applied AI, Fidji Simo, stated during a Thursday press briefing that the company aims to "unlock more economic value for people" with this release. She noted that the model had been in development for "many, many months."
OpenAI described GPT-5.2 in its press release as its "most suitable model for daily professional use to date," clearly positioning it against Gemini 3, currently regarded as the leading general-purpose model. The company highlighted that the GPT-5.2 series—including Instant, Thinking, and Pro versions—performs better in tasks such as "spreadsheet creation, presentation building, coding, image comprehension, long-context processing, tool usage, and handling complex multi-step tasks."
Aidan Clark, OpenAI's VP of Research, shared that the team had granted early access to GPT-5.2 Pro to a senior immunology researcher. When asked to generate "the most important unsolved questions in the immune system," the researcher reportedly found that GPT-5.2 provided "sharper questions" and "more compelling explanations" about their significance compared to other cutting-edge models.
In a blog post, OpenAI also emphasized GPT-5.2's superior performance in supporting AI agent workflows—a key battleground for major AI firms. Researcher Max Schwarzer stated during the briefing that the company's vision is to make ChatGPT "the best possible personalized assistant."
OpenAI stressed that its Thinking model exhibits "fewer hallucinations" than its predecessor, enhancing its reliability as a trusted agent in professional settings. Early testers, including Notion, Box, Shopify, Harvey, Zoom, and Databricks, gained access weeks ago.
The release of GPT-5.2 comes less than a month after GPT-5.1, which focused on improving the model's "temperature control" and "chat experience," including preset personalities for tone customization. OpenAI claims GPT-5.2 further refines these aspects: "Users will find GPT-5.2 more structured and reliable in daily use while remaining pleasant and easy to interact with."
OpenAI is also competing with Google's rapid advancements. Reports indicate the company plans to launch two new models next year, one codenamed "Garlic." Additionally, CEO Sam Altman reportedly issued an internal "red alert" memo last week, delaying certain initiatives (like advertising) to prioritize ChatGPT's capabilities. Simo declined to share updates on the advertising timeline during the briefing.
"We announced this 'red alert' to signal our focus on concentrating resources in this specific area," Simo explained.
On the same day as GPT-5.2's launch, another major announcement emerged: OpenAI has secured a three-year licensing deal with Walt Disney, permitting the creation of "user-prompted social videos" featuring over 200 characters from Disney, Marvel, Pixar, and Star Wars. As part of the agreement, Disney will host select videos on Disney+ while becoming a "major customer" of OpenAI and making a $1 billion equity investment in the company.
OpenAI also mentioned in its blog that it's in the "early stages" of developing an age-prediction model to automatically apply additional safety measures for minors. Simo revealed testing has begun in select countries to ensure the model identifies teens "without misclassifying adults," with hopes to complete trials before launching ChatGPT's "adult mode." The feature is expected to debut in Q1 2026.
GPT-5.2 will roll out today within ChatGPT, initially for paid users (Plus, Pro, Go, Business, and Enterprise). OpenAI plans a "gradual deployment" to ensure a "smooth and reliable" chat experience. The previous GPT-5.1 model will remain available under the "legacy models" menu for three months before being phased out.
Comments