OpenAI is rushing to release its GPT-5.2 model as soon as next week, responding to CEO Sam Altman's declaration of a "code red" situation earlier this week as the company faces intensifying competition from Google's Gemini 3 and other artificial intelligence (AI) rivals, according to recent reports.
AI Generated Image
The Verge reported on Friday that sources familiar with OpenAI's plans say the GPT-5.2 update, originally scheduled for later this month, has been moved forward to December 9th. The accelerated release aims to close the performance gap created when Google launched Gemini 3 last month, a model that topped industry leaderboards.
Altman told employees in an internal memo Monday that OpenAI would delay initiatives including advertising and AI agent development to marshal resources for improving ChatGPT, according to the Information earlier this week. The move marks a role reversal from three years ago when Google launched its own code red to respond to ChatGPT's threat to Google Search.
The urgency comes as OpenAI estimates ChatGPT handles 70% of global AI assistant activity, but faces growing pressure from Google's Gemini, which reported 650 million monthly active users in October. ChatGPT's performance is critical for OpenAI's ability to raise another $100 billioin to cover projected cash burn while pursuing revenue targets of $10 billion this year and $35 billion by 2027.
Imminent Model Update to Counter Google
GPT-5.2 is ready for deployment and should narrow the advantage Google established with Gemini 3, sources told The Verge. According to the Information, Altman informed colleagues that OpenAI's next reasoning model performs "ahead of Gemini 3" in the company's internal evaluations. Reasoning models use additional computing power to generate superior responses, powering ChatGPT features like Thinking mode and Deep Research.
However, OpenAI's planned release dates frequently shift due to development challenges, server capacity constraints, or rival announcements, meaning GPT-5.2 could arrive slightly later than December 9th. Warren noted that the company prioritizes responding to competitive pressures over adhering to fixed schedules.
'Code Red' Refocuses Resources on Core Product
Altman's Monday memo detailed that the code red surge would delay projects including advertising features OpenAI has been testing for online shopping queries, AI agents for automating shopping and health tasks, and Pulse, a feature generating personalized morning reports for users.
The CEO didn't specify ChatGPT's exact problems, but Google stated this fall that Gemini had gained usage ground. CFO Sarah Friar alluded to a ChatGPT growth slowdown during a call with investors last month, according to a person with knowledge of her remarks, though the specific metric remained unclear.
Altman identified key priorities including personalizing ChatGPT for over 800 million weekly users, improving Imagegen image-generation capabilities following Google's Nano Banana Pro release, enhancing model behavior to win public rankings like LMArena, boosting speed and reliability, and reducing overrefusals when the chatbot unnecessarily declines benign questions.
'Garlic' Model Under Development with Pretraining Advances
Beyond the immediate GPT-5.2 response, OpenAI is developing a larger language model codenamed Garlic that may represent a more significant competitive advance, the Information reported Tuesday. Chief research officer Mark Chen told colleagues last week that Garlic was performing well on company evaluations against Gemini 3 and Anthropic's Opus 4.5 in coding and reasoning tasks, according to a person with knowledge of his remarks.
Chen said OpenAI aims to release a Garlic version as soon as possible, potentially as GPT-5.2 or GPT-5.5 by early next year. The model incorporates pretraining improvements—the initial training stage where models learn from web data—addressing problems that limited GPT-4.5, which launched in February. Chen said the advances allow OpenAI to infuse smaller models with knowledge previously requiring much larger, costlier models.
Garlic differs from Shallotpeat, another model Altman mentioned in October for challenging Gemini 3. Google acknowledged making pretraining breakthroughs with Gemini 3, and OpenAI leaders have recognized the critical importance of such fundamental advances. Garlic still requires post-training with curated data for specific fields, plus testing and safety evaluations before release.
Explore more exclusive insights at nextfin.ai.
