Google Nears Release of Flagship Gemini 4 AI Model

Google is nearing the release of its newest flagship model, Gemini 4, the head of its DeepMind division said Wednesday, a long-awaited development after the tech giant fell behind rivals Anthropic and OpenAI in the AI model race.

Speaking at The Information’s AI Agenda Live Summit, in his first media appearance as the new top leader of Google DeepMind, Koray Kavukcuoglu said Gemini 4 is in the early days of a phase of development called post-training, in which a base AI model is refined to behave reliably. He said he hopes the model will be released “much earlier” than the end of the year.

“Our intention is to, like—as soon as possible—to release an early post-training output because we see the results and we are excited,” he said. Then, Google would “continue the fast-paced iterations.”

Gemini 4 would be Google’s first next-generation flagship model since the launch of the Gemini 3 line late last year—though it released an update, Gemini 3.1, in February and has released a series of smaller, cheaper Flash models less capable of complex reasoning. Google CEO Sundar Pichai had said in May at the company’s annual I/O conference that it would release a major update, Gemini 3.5 Pro, in June, but the release never came.

Kavukcuoglu said that instead of releasing Gemini 3.5 Pro, the company “took a little bit of a step back” after the release of Gemini models 3 and 3.1 and focused on introducing the Flash models.

“For us, what was important is to maximize the speed of learning at the time,” he said. He didn’t specify whether Google has officially scrapped Gemini 3.5 Pro, but made clear its focus is releasing Gemini 4.

Rival Anthropic this spring launched its powerful new Mythos family of models, which it upgraded this month, and OpenAI started launching its similarly capable GPT-6 models in early September. Both are more capable than Google’s most advanced current models.

Asked Wednesday whether he worries that Google has fallen behind peers, Kavukcuoglu said he’s “a hundred percent confident” in his team.

“We have [an] insanely strong team, and I have the utmost trust in the team,” he said. “In my mind, it’s a certainty that we are always gonna be at the frontier.”

Kavukcuoglu said Google is in the middle of developing guardrails and safety testing for Gemini 4, which is being used internally by engineers to run its Antigravity AI coding tool.

This story is developing and will be updated.