Google DeepMind chief Koray Kavukcuoglu announced that the lab is prioritizing an early release of its upcoming flagship model, Gemini 4, before the end of the year. Speaking at The Information’s AI Agenda Live Summit in his first media appearance since taking over leadership from Demis Hassabis in August, Kavukcuoglu stated the model is currently undergoing early post-training and internal developer testing.
The push for an expedited rollout follows internal delays; the previously announced Gemini 3.5 Pro missed multiple deadlines and failed to ship after being scheduled for June. In the interim, major competitors Anthropic and OpenAI launched updated model suites, including Anthropic’s Mythos/Fable lineup and OpenAI’s GPT-6 releases, putting commercial pressure on Google’s AI unit.
In a notable strategic pivot from his predecessor, Kavukcuoglu signaled a shift away from public framing around achieving artificial general intelligence (AGI), calling it “not the right conversation.” Instead, DeepMind is focusing on practical reliability and deploying actionable, trustworthy intelligent agents while utilizing new model developments to inform Google’s custom TPU silicon roadmap.
Why it matters
Signals a commercial realignment at Google DeepMind from theoretical AGI research toward faster release cycles and enterprise utility.
Highlights intense competition among frontier labs to deliver updated flagship models as rivals roll out newer architectures.
Demonstrates tight integration between model development and custom AI silicon (TPUs) to optimize hardware design.
Source: the-decoder.com



