Google DeepMind’s new leader, Koray Kavukcuoglu, says Gemini 4 is in its refinement stage and that Google aims to release it “much earlier” than the end of 2026. He gave that update in an interview with The Information, as reported by The Verge. Google has given no launch date in the report.
Kavukcuoglu said Google wants to release an early output from post-training as soon as it can, then keep iterating. That describes a possible release sequence, but it does not establish which Gemini 4 model will arrive first or what developers will be able to use at launch.
The gap between a timeline hint and a model release
The Verge reports that Google last released a flagship Gemini model with the Gemini 3 series in November 2025. A Gemini 3.5 Pro update that CEO Sundar Pichai mentioned for June did not roll out. Kavukcuoglu said the team stepped back from that update to focus on faster Flash models.
That history makes the new timing comment useful, but limited. “Much earlier” than year-end offers a direction, not a month. The reported refinement stage says work is ongoing. It gives no independent performance result, API name, price, context limit, or availability date.
For teams choosing a model this quarter, the operational question is when Gemini 4 can be tested on their own tasks. The answer depends on a release announcement and access details that Google has yet to provide in the reported interview.
