Koray Kavukcuoglu, the new head of Google DeepMind, confirmed that the Gemini 4 model is currently in a refinement phase with a target launch date moving earlier than the original end-of-year schedule. Speaking to The Information during his initial media appearance, Kavukcuoglu stated the team intends to release an early post-training output as soon as possible after observing internal results. This announcement addresses previous delays where Google trailed rival developers on flagship AI releases while the company focused on stabilising the model. The shift suggests a move away from waiting for perfect alignment to prioritising rapid iteration and early market availability. Kavukcuoglu emphasised that the fast-paced approach allows the team to see real-world performance metrics quickly rather than waiting for finalised documentation. This strategy aims to reduce the gap between development and deployment while maintaining the core capabilities expected from the next generation of the system.
- The model is currently undergoing refinement rather than full deployment testing.
- Google plans to release an early version before the year concludes.
- Kavukcuoglu will lead the DeepMind division from this point forward.
