On December 2, 2025, The Information reported that OpenAI is in the process of developing a new large language model, internally codenamed 'Garlic'. This move is aimed at taking on Google's Gemini 3 in the competitive landscape. In tests focusing on programming and logical reasoning tasks, 'Garlic' has outperformed both Google's Gemini 3.0 and Anthropic's Opus 4.5. The key advancement behind this achievement is the optimization of the algorithm architecture during the pre-training stage. This optimization allows a model with a relatively small number of parameters to achieve a high knowledge density. As a result, it substantially cuts down on both training costs and time.

In response to this development, OpenAI CEO Sam Altman has activated an emergency response mechanism. The primary goal is to boost the performance of ChatGPT. The plan is to first roll out a new model that has been optimized for reasoning capabilities. Looking ahead, 'Garlic' is anticipated to be released in early 2026, potentially under the names GPT-5.2 or GPT-5.5. Moreover, OpenAI has already embarked on the development of an even larger-scale model.