Google DeepMind Executive Says Gemini 4 Model Advancing in Training Phase

What Factors Are Influencing the Accelerated Timeline for Gemini 4?

Koray Kavukcuoglu, senior vice president at Google DeepMind, confirmed that work on Gemini 4 is underway and progressing through its initial post-training stages. Speaking in an interview shared via Techmeme, he expressed optimism that the model could be released significantly sooner than previously anticipated, potentially well before the end of 2026. The update comes as Google continues to expand its AI capabilities amid growing competition in the large language model space.

Kavukcuoglu emphasized that Gemini 4 is currently in the early phase of post-training, a critical stage where models are refined using specialized datasets to improve He noted that the team is focused on enhancing the model’s ability to handle complex tasks across multiple modalities, building on the foundation laid by earlier versions. While he did not disclose specific technical benchmarks, his comments suggest confidence in the development trajectory and a desire to accelerate the release timeline.

How Will Gemini 4 Build on Previous Generations?

The push to release Gemini 4 „much earlier” than the end of the year appears driven by internal milestones and external competitive pressures. Kavukcuoglu indicated that recent advances in training efficiency and data curation have allowed the team to move faster than expected. Additionally, the increasing demand for more capable AI systems in enterprise and consumer applications may be motivating Google to shorten its release cycle. He stressed that speed would not come at the expense of safety or performance, with rigorous testing remaining a priority throughout development.

Gemini 4 is expected to expand upon the multimodal strengths of its predecessors, integrating improved understanding of text, images, audio, and video. Kavukcuoglu highlighted that post-training efforts are particularly aimed at refining contextual awareness and reducing hallucinations in nuanced scenarios. The model may also feature enhanced long-context handling and better integration with Google’s ecosystem of products, including Search, Workspace, and Cloud services. These upgrades aim to position Gemini 4 as a more reliable and versatile tool for developers and end-users alike.

Frequently Asked Questions

When is Gemini 4 expected to be released? Kavukcuoglu hopes the model will be released „much earlier” than the end of 2026, though no exact date has been provided.

What stage of development is Gemini 4 currently in? The model is in the early phase of post-training, where it undergoes refinement to improve accuracy and Will Gemini 4 include new capabilities compared to Gemini 3? Yes, it is expected to advance multimodal