OpenAI’s New ‘Garlic’ AI Model: Can It Keep Up With Google?

OpenAI is reportedly working on a new large language model codenamed “Garlic,” a move that signals just how quickly the competitive balance in AI can shift. 

According to reporting based on The Information, OpenAI’s Chief Research Officer Mark Chen told colleagues the model has been testing strongly inside the company, especially on coding and reasoning tasks, where it was described as performing well against Google’s Gemini 3 and Anthropic’s Opus 4.5. (Bakolia; “OpenAI Develops New Model”; Tech Desk)

That urgency is showing up not just in what OpenAI is building, but in how it is reorganizing. CEO Sam Altman has reportedly declared a “code red,” urging teams to prioritize improving ChatGPT’s quality and, in an internal memo, to delay other initiatives including broader advertising plans. The focus, according to reports, is on making ChatGPT more responsive and better personalized as rivals push forward with rapid releases and ecosystem rollouts. (Bakolia; Tech Desk)

The timing matters. Google released Gemini 3 on November 18 and moved quickly to integrate it across its wider product ecosystem, which has contributed to the sense that Google is executing its fastest rollout yet. Reports describe Gemini 3 as posting top benchmark results across image editing, text generation, and multimodal reasoning, while Anthropic positioned Claude Opus 4.5 as the “world’s best” model for coding. Against that backdrop, OpenAI is also preparing a new reasoning model that Altman reportedly told colleagues is “ahead” of Gemini 3 in internal evaluations, underscoring how central coding, reasoning, and reliability have become in the race for both users and enterprise credibility. (Tech Desk; “OpenAI Develops New Model”)

Garlic is also portrayed as more than just a competitive patch. Chen reportedly framed it as the result of solving key pretraining issues, including improvements beyond OpenAI’s “previous best” and larger pretrained model, GPT-4.5, which launched in February but has since faded from view. In practical terms, the reported breakthrough is that OpenAI can now “infuse” a smaller model with an amount of knowledge that previously required training a much larger model, a meaningful shift because bigger models are typically slower and more expensive to train. Garlic is also described as distinct from another in-progress model called Shallotpeat, which Altman reportedly mentioned internally in October as part of OpenAI’s strategy to challenge Gemini 3. Notably, Garlic is said to include bug fixes discovered while developing Shallotpeat, and those fixes were incorporated during pretraining, the stage where models learn broad patterns from large-scale data before being refined with more curated information. (Bakolia)

Even with strong internal results, Garlic still has steps to clear before any public launch. Reports say it must go through post-training with more carefully selected data, additional testing, and safety evaluations. If released soon, Garlic could reportedly arrive as GPT-5.2 or GPT-5.5 by early 2026, and some coverage suggests OpenAI is positioning itself to deliver more specialized, high-value applications, including in industries like biomedicine and healthcare. Meanwhile, pressure is not only technical but commercial: one report cites entrepreneur Deedy Das claiming that in the two weeks after Gemini 3’s launch, ChatGPT’s unique daily active users (7-day average) fell by 6 percent, a worrying signal for a company whose revenue depends heavily on consumer momentum and whose reported $500 billion valuation assumes continued growth toward a $20 billion annual recurring revenue target. (Bakolia; “OpenAI Develops New Model”; Tech Desk)

This article was written by the Covertly.AI team. Covertly.AI is a secure, anonymous AI chat that protects your privacy. Connect to advanced AI models without tracking, logging, or exposure of your data. Whether you’re an individual who values privacy or a business seeking enterprise-grade data protection, Covertly.AI helps you stay secure and anonymous when using AI. With Covertly.AI, you get seamless access to all popular large language models - without compromising your identity or data privacy.

Try Covertly.AI today for free at www.covertly.ai, or contact us to learn more about custom privacy and security solutions for your business.  

Works Cited

Bakolia, Ravikash. “OpenAI developing new AI model ‘Garlic’ amid Google’s AI success: report.” Seeking Alpha, 2 Dec. 2025, https://seekingalpha.com/news/4527661-openai-developing-new-ai-model-garlic-amid-googles-ai-success.

“OpenAI develops new model codenamed Garlic to compete with Google - The Information.” Investing.com, 2 Dec. 2025, https://www.investing.com/news/stock-market-news/openai-develops-new-model-codenamed-garlic-to-compete-with-google--the-information-93CH-4386486.

Tech Desk. “OpenAI developing new LLM codenamed ‘Garlic’ as ChatGPT goes into code red mode.” The Indian Express, 3 Dec. 2025, https://indianexpress.com/article/technology/artificial-intelligence/openai-developing-new-llm-codenamed-garlic-chatgpt-code-red-10399404/.

Garg, Ankita. “ChatGPT Outage Fixed After Brief Glitch as OpenAI Hits Code Red Amid Rising Google Rivalry.” India Today, 3 Dec. 2025, https://www.indiatoday.in/technology/news/story/chatgpt-outage-fixed-after-brief-glitch-as-openai-hits-code-red-amid-rising-google-rivalry-2829934-2025-12-03. 

Seb. “What’s the Deal with OpenAI’s Garlic Model?” Let’s Just Do AI, 2 Dec. 2025, https://letsjustdoai.com/whats-the-deal-with-openais-garlic-model/. 

Kahan, Raphael. “OpenAI Declares ‘Code Red’ as Google’s Gemini Closes the ChatGPT Gap.” Ynetnews, n.d., https://www.ynetnews.com/tech-and-digital/article/h1kkevh1111x.