New DeepMind Chief Says Gemini 4 Could Launch Earlier Than Expected

Google DeepMind is moving closer to the release of Gemini 4, with its newly appointed chief, Koray Kavukcuoglu, revealing that the next-generation AI model has entered the early stages of post-training. This stage involves refining a trained model to improve its reliability, behaviour and performance before it is released to the public. Kavukcuoglu said Google wants to release an early post-training version as soon as possible because the results so far have been encouraging.

The comments were made during The Information’s AI Agenda Live Summit, in Kavukcuoglu’s first major media appearance since taking over leadership of Google DeepMind. He indicated that Gemini 4 could arrive much earlier than the end of 2026, although Google has not announced a specific public launch date. The company plans to continue rapidly improving the model after its initial release through further iterations.

Gemini 4 is expected to become Google’s next major flagship AI system following the Gemini 3 generation. Google CEO Sundar Pichai previously confirmed that the company had begun its most ambitious pre-training effort yet for Gemini 4. The decision to concentrate on Gemini 4 followed Google’s shift away from an expected Gemini 3.5 Pro release toward smaller Flash models, with the company prioritising faster development and learning.

The new model is being developed amid intense competition from OpenAI and Anthropic, which have continued releasing increasingly capable AI systems. Kavukcuoglu acknowledged the competitive environment but expressed confidence in Google’s research team. He also said Google is working on safety testing and guardrails for Gemini 4, which is already being used internally by engineers in connection with Google’s Antigravity AI coding tool.

Gemini 4 is also important to Google’s wider hardware strategy. Kavukcuoglu said the development of frontier AI models gives Google’s hardware teams information that can influence the design of future generations of its Tensor Processing Units (TPUs). Google has increasingly expanded its TPU business as it competes in the AI-computing market, meaning advances in Gemini could influence not only Google’s software products but also the company’s future AI-chip development.