Google is nearing the launch of its Gemini 4 model, with DeepMind leader Koray Kavukcuoglu saying the frontier AI system is in its refinement stage. The company aims to release the model "much earlier" than the end of 2026, Kavukcuoglu said during his first media appearance as head of Google's DeepMind division at The Information's AI Agenda Live Summit on Sept. 23-24.
Google has lagged behind rival developers in releasing flagship AI models. The company has not introduced a new top-tier AI model since the Gemini 3 series launched in November 2025. OpenAI's GPT-6 and Anthropic's Mythos series models have since outperformed Google's Gemini 3, giving those providers a notable lead.
Google CEO Sundar Pichai announced a major Gemini 3.5 Pro update at the I/O event in May, but that model was never rolled out to users. Kavukcuoglu explained that Google "took a little bit of a step back" from that release, instead shifting focus to introducing faster, less powerful Flash models.
Former DeepMind AI head Demis Hassabis stepped down in August after stating that "great progress" was being made on Gemini 4.
Kavukcuoglu expressed confidence in Google's long-term position. "In my mind, it's a certainty that we are always gonna be at the frontier," he said. Google intends to release an early post-training output as soon as possible, with Kavukcuoglu saying the company is excited by current results and will "continue the fast-paced iterations" for future development.
