Koray Kavukcuoglu wants to release Gemini 4 well before the end of the year. He calls the AGI question, which his predecessor Demis Hassabis always focused on, "not the right conversation." Under pressure from OpenAI and Anthropic, Google's former research lab has become a Gemini factory.
Kavukcuoglu made his first media appearance as head of Google Deepmind at The Information's AI Agenda Live Summit. He said Gemini 4 is in an early phase of post-training, the stage where a base model gets refined to behave reliably. He hopes to release it "much earlier" than the end of the year.
"Our intention is to [...] as soon as possible [...] release an early post-training output because we see the results and we are excited," Kavukcuoglu said. Fast iterations would follow.
Google is still working on safeguards and safety testing. Engineers are already using the model internally for the coding tool Antigravity.
Kavukcuoglu, previously CTO, took over in August after Demis Hassabis stepped down as Deepmind CEO. Hassabis now chairs the unit and serves as Alphabet's chief scientist.
Google has ground to make up after Gemini 3.5 Pro quietly vanished
Gemini 4 would be Google's first new flagship model since Gemini 3 late last year. Since then, the company has shipped the Gemini 3.1 update in February and several smaller, cheaper Flash models. CEO Sundar Pichai announced Gemini 3.5 Pro at I/O in May for a June release, but it never arrived. According to the Financial Times, the project missed three deadlines.
Kavukcuoglu said Google took "a little bit of a step back." "For us, what was important is to maximize the speed of learning at the time," he explained. He didn't say whether Gemini 3.5 Pro has been officially canceled.
Rivals have moved ahead in the meantime. Anthropic released its Mythos/Fable family in the spring and updated it, along with the Opus line, this month. OpenAI has been rolling out GPT-6 models since early September. All of them are far more capable than Google's current top models.
Kavukcuoglu says he isn't worried. He's "a hundred percent confident" and trusts his "insanely strong team." "In my mind, it's a certainty that we are always gonna be at the frontier," he said.
The model work is also supposed to help Google's chip business. The company started selling its TPUs directly to customers this year, putting it in closer competition with Nvidia. According to Kavukcuoglu, new models show hardware engineers how to design the "next two or three generations" of chips.
Google has caught up before. Gemini 3 was such a leap that OpenAI declared a "Code Red" and only pulled even again with GPT-5.2. Whether Gemini 4 can repeat that remains an open question.
Deepmind drifts from its AGI mission as its top researchers walk out
What the new chief says about AGI, long Deepmind's most important goal, is just as revealing as the timeline. "The conversation of [whether] we achieved AGI or not is not the right conversation," Kavukcuoglu said. "The conversation is more about are we able to build intelligent agents that we can trust?"
That's a clear signal from a lab that long defined itself by the pursuit of artificial general intelligence. Just a few weeks ago, his predecessor sounded very different. "I've been working towards AGI my whole life and now, like many of you, I feel it is close at hand," Hassabis wrote in his farewell memo.
Hassabis co-founded Deepmind in 2010, and Google acquired the lab four years later. The London team enjoyed special status inside the company for years. ChatGPT changed that. During Google's first "Code Red," the company merged Google Brain and Deepmind into "Google Deepmind," which is also where the Gemini models got their name.
The August restructuring caps that shift for now. Kavukcuoglu runs day-to-day operations without the CEO title, and all Gemini development is moving to the Bay Area. "The era of DeepMind as an independent actor is over," a former Google executive told the Guardian. According to the Financial Times, executives were frustrated that Hassabis gave commercial needs less weight, for example when he decided to release AlphaFold openly.
It's unclear how voluntary Hassabis' exit really was. Citing industry sources, the newsletter Pathfounders reports that Hassabis wanted to leave alongside Jeff Dean, but management feared the stock price would crash. Google hasn't commented on the reports.
The talent drain, on the other hand, isn't in dispute. Noam Shazeer left for OpenAI in June, and two days later AlphaFold Nobel laureate John Jumper announced he was joining Anthropic. Dean founded the startup Discovery Loop with Oriol Vinyals, Quoc Le, and Sanjay Ghemawat. Gemini researchers Jonas Adler and Alexander Pritzel also left for Anthropic. According to Fortune, Deepmind's ratio of new hires to departures fell from about 12 to 1 in the second quarter of 2023 to roughly 2 to 1 in the third quarter of 2026.
Under Kavukcuoglu, Deepmind's shift to a product unit is complete
The timing stands out too, because the AGI debate is louder now than it's been in a long time. On September 12, Anthropic CEO Dario Amodei published his essay "We Must Pace the Frontier." In it, he called on frontier labs to slow how quickly they push their models' capabilities, without halting development altogether. OpenAI followed suit, Elon Musk agreed, and Hassabis said the direction was right but the details still needed work. Back in July, more than 1,200 employees at frontier labs had signed a letter with the same name.
Kavukcuoglu isn't brushing off safety, since trustworthy agents are a safety issue in their own right. But while his predecessor and Google's rivals argue about AGI and its risks, the new Deepmind chief treats progress as a product question. The pressure ChatGPT set off, which OpenAI and increasingly Anthropic keep up, has turned a research lab with an AGI mission into Google's Gemini factory. That shift now looks complete.