Koray Kavukcuoglu, head of Google's DeepMind, stated that Gemini 4 is currently in the refinement stage, with Google aiming to launch the model well before the end of this year. Since the release of the Gemini 3 series in November 2025, Google has not introduced a new flagship AI model.
This marks Kavukcuoglu's first media interview since assuming the role of DeepMind head. He said the team hopes to release early results from the post-training phase as soon as possible, followed by continued rapid iterations. Currently, Google has provided a launch target rather than a confirmed release date.
Following the release of the Gemini 3 series, OpenAI launched GPT-6, and Anthropic also introduced its Mythos series of models. DeepMind previously underwent leadership changes: Demis Hassabis stepped down from his relevant position this August, stating before his departure that Gemini 4 had made "significant progress."
Google CEO Sundar Pichai had stated at the I/O event this May that Gemini 3.5 Pro would receive a major update in June, but the model was not released as initially announced. Kavukcuoglu explained that Google subsequently shifted its focus to the faster, less capable Flash models. When asked if he was concerned about Google falling behind competitors, he expressed confidence that Google would always remain at the forefront of AI technology.