Breaking: OpenAI’s new model is codenamed “Garlic”. Internal benchmarks show it beating Gemini 3 and Opus 4.5.

OpenAI Announces New Language Model “Garlic” Aims to Surpass Competitors in Performance

In a recent development within the artificial intelligence research community, OpenAI has unveiled a new model internally codenamed “Garlic.” Preliminary internal benchmarks indicate that this model surpasses notable competitors such as Google’s Gemini 3 and Anthropic’s Opus 4.5 in key areas of coding proficiency and reasoning capabilities.

Innovations and Architectural Advancements

“Garlic” addresses several structural limitations identified in OpenAI’s earlier project, colloquially referred to as Shallotpeat. The new model achieves a remarkable balance by integrating extensive “big model knowledge” within a more streamlined and efficient architectural framework. This approach allows for leveraging extensive training data and sophisticated modeling techniques without the need for an excessively large parameter count.

Competitive Edge and Performance Metrics

According to internal assessments, “Garlic” demonstrates superior performance in both coding tasks and complex reasoning evaluations—areas critical to the evolving demands of AI applications. These early results suggest that the model could redefine OpenAI’s position in the competitive landscape of large language models.

Projected Timeline and Industry Implications

OpenAI is reportedly aiming for a formal release of “Garlic” around early 2026, possibly under the GPT-5.2 or GPT-5.5 branding. The goal is to re-establish dominance ahead of rivals and to set new standards in AI capabilities.

Significance of the “Garlic” Codename and Broader Market Context

The choice of the codename “Garlic” may symbolize a strategic move—perhaps signaling strength, versatility, or a new flavor in AI technology. While the competitive landscape, notably Google’s recent advancements, presents challenges, OpenAI’s ongoing innovations might prove pivotal in maintaining its leadership position.

Conclusion

As the AI community eagerly anticipates the official launch, the development of “Garlic” underscores OpenAI’s commitment to pushing the boundaries of language model performance. Its potential to outperform current leading models could have significant implications across various fields, from software development to complex reasoning applications.

For more detailed insights, refer to the original report by The Information: Link to Source