THE CRUNCH
Google DeepMind's new chief, Koray Kavukcuoglu, has confirmed that the long-awaited Gemini 4 model is nearing launch. Speaking to The Information, Kavukcuoglu stated that the model is currently in its refinement stage and that Google intends to release an early post-training output as soon as possible. He indicated the launch will happen much earlier than the end of the year, marking a shift in Google's pace on its旗舰
Kavukcuoglu emphasised that the team is excited by the results and intends to continue the fast-paced iteration cycle that has characterised recent developments. This suggests Google is moving away from the extended delays that have previously characterised its flagship release cadence.
The announcement comes as Google seeks to close the gap with rival developers who have been releasing their own flagship models. The focus on an early release of post-training outputs implies a strategy to quickly iterate on performance and capabilities.
Google DeepMind researchers have previously explored methods like Dream-RSI, which lets AI agents improve search strategies by replaying past attempts. This work on computational efficiency and self-improvement may inform the practical deployment of future models like Gemini 4.