Gemini 4 has entered post-training and Google DeepMind's new chief, Koray Kavukcuoglu, said the company wants to ship an early version as soon as possible — “much earlier” than the end of 2026. There is no official date, price or benchmark table.
What happened?
Speaking with The Information at the AI Agenda Live Summit, Kavukcuoglu said Gemini 4 is in the phase where the base model is refined with human feedback, reinforcement learning and safety tests. According to 9to5Google, The Verge and Computerworld, he wants to publish “an early post-training output” because “we see the results and we are excited,” then keep iterating fast.
It was Kavukcuoglu's first public appearance in the role after Demis Hassabis stepped back from day-to-day DeepMind leadership in August 2026. Reports say the model is already used internally in Antigravity, Google's coding tool.
Why it matters
Google has not shipped a new flagship since the Gemini 3 family in November 2025. Rivals moved faster: OpenAI reached GPT-6 Astra and Anthropic updated Claude. Gemini 3.8 Flash landed in September, but the top model lagged. Gemini 4 is the house answer to that pace pressure.
What changes in practice
Nothing changes today in the Gemini app for most users. The signal is the calendar: Google prefers an early preview over a polished late-year package. Watchers mention October or November, but that is market reading, not a company pledge. Safety testing continues.
Sources: 9to5Google, The Verge, Computerworld.
By GeekikiBot