Gemini 4 is almost ready, says new Google DeepMind chief

The update comes directly from Koray Kavukcuoglu, who spoke with The Information in what marks his very first media appearance as the newly appointed leader of Google’s prominent DeepMind division. During the interview, Kavukcuoglu revealed that the Gemini 4 model is currently undergoing its final refinement stage. He indicated that the company is actively aiming to bring the model to the market "much earlier" than the originally anticipated late-year timeline.

"Our intention is to, like — as soon as possible — to release an early post-training output because we see the results and we are excited," Kavukcuoglu told the publication. He further emphasized that Google intends to "continue the fast-paced iterations" moving forward.

This aggressive stance hopefully signals a return to a more consistent and reliable AI release cadence for Google. The company has recently experienced notable shifts and restructuring at the highest echelons of DeepMind’s leadership. Most prominently, former AI head Demis Hassabis stepped down from his role in August. Prior to his departure, Hassabis noted that "great progress" was actively being made on the development of the upcoming Gemini models, specifically highlighting Gemini 4.

Google’s last major flagship AI release was the Gemini 3 series, which rolled out to the public in November 2025. In the months following that release, the artificial intelligence landscape shifted rapidly as competitors moved ahead with major hardware and model updates. OpenAI introduced its GPT-6 models, while Anthropic rolled out its Mythos series. Both of these competitive offerings have outperformed the Gemini 3 series in various benchmarks and practical evaluations, handing rival AI providers a sizable and noticeable lead against Google in the broader, ongoing artificial intelligence race.

The pressure on Google to respond has been mounting throughout the year. During the annual Google I/O developer event in May, Google CEO Sundar Pichai announced that a major Gemini 3.5 Pro update would be coming down the pipeline in June. The stated goal of this update was to help bridge the widening performance gap between Google’s existing models and those of its competitors. However, the expected rollout faced delays and the Gemini 3.5 Pro model was ultimately never released to the public, leaving users and industry analysts waiting for a substantive counterpunch.

Addressing the delay and the shifting strategy, Kavukcuoglu explained that the company deliberately "took a little bit of a step back" from its original plans. Instead of pushing forward with the Gemini 3.5 Pro release, Google redirected its engineering and research focus toward introducing faster, albeit less powerful, Flash models to satisfy immediate user and enterprise demands for speed and efficiency.

When pressed on whether he harbors concerns about Google potentially falling behind its chief rivals in the hyper-competitive field of artificial intelligence development, Kavukcuoglu remained steadfast in his confidence. Speaking with The Information, he dismissed the notion of long-term lagging, asserting that "In my mind, it’s a certainty that we are always gonna be at the frontier."