Google DeepMind's new chief confirms Gemini 4 in post-training, targets release 'far earlier' than year end

On September 23, Google's new DeepMind head Koray Kavukcuoglu made his first public appearance in the role at The Information's AI Agenda Live summit, confirming that the company's next flagship Gemini 4 is in early post-training and is targeted for release 'far earlier' than year end.

Kavukcuoglu said: 'Our intention is, as soon as possible, to release an early post-training output because we see the results and we are excited.' He added that Google will continue iterating rapidly after the first cut. Gemini 4 will be Google's first next-generation flagship since the Gemini 3 series launched at the end of 2025. In the intervening period, OpenAI shipped its GPT-6 family and Anthropic introduced Mythos-class models, both of which outperform Gemini 3 on the current frontier.

Kavukcuoglu noted that after shipping Gemini 3 and 3.1, Google did not proceed with Gemini 3.5 Pro but 'took a little bit of a step back' to focus on Flash models. The focus has now shifted decisively to Gemini 4. Asked whether Google is falling behind, he expressed full confidence in the team, saying, 'In my mind, it's a certainty that we are always going to be at the frontier.' Gemini 4 is currently in safeguard development and safety testing; engineers are already using it internally to run the Antigravity AI coding tool.