Koray Kavukcuoglu, the newly appointed leader of Google’s DeepMind division, has indicated that the company is approaching the release of its next flagship AI model, Gemini 4. In a recent discussion with The Information, Kavukcuoglu confirmed that the model has reached its refinement stage. This development follows a period where Google was perceived as trailing behind other major AI developers in the release of top-tier foundation models. The Verge reported on the executive’s first media appearance in his new role, highlighting the shift toward finalized testing for the upcoming system.
The Shift in Google’s AI Strategy
The transition to Gemini 4 represents a critical step for Google as it seeks to regain momentum in the competitive landscape of generative artificial intelligence. For much of the past year, industry observers have noted that Google appeared to move at a more cautious pace compared to rivals. By moving Gemini 4 into the refinement stage, the company signals that its core architecture is stable and is now undergoing the fine-tuning necessary for public or enterprise deployment. This phase typically involves safety testing, performance optimization, and alignment to ensure the model adheres to internal quality standards before a wide-scale rollout.
Technical Details and Launch Timeline
While the confirmation of the model’s progress is significant, the announcement did not specify a concrete release date for Gemini 4. It remains unknown whether the model will be released in stages or if it will be available to developers and consumers simultaneously. Furthermore, Kavukcuoglu did not provide technical specifications regarding the model’s parameter count, its performance on industry benchmarks, or its specific capabilities compared to previous versions of the Gemini family. The announcement also did not specify the hardware requirements for hosting the model or the pricing structure for API access. It is unclear if Gemini 4 will introduce new modalities or if it focuses primarily on improving the reasoning and speed of existing text and vision tasks.
As Gemini 4 moves through its final stages of development, the industry will be watching to see if this release can bridge the gap with its primary competitors. Google’s ability to translate this refinement stage into a successful launch will be a key test for Kavukcuoglu’s leadership at DeepMind.
Lead image: panumas nikhomkhai / Pexels
