Gemini 3 has beaten competitors and made it to the top of most benchmarks, but its creators say that AGI is still some way away.
In a candid assessment of Google DeepMind’s progress toward artificial general intelligence, CEO Demis Hassabis struck a balance between satisfaction and caution. Speaking about the company’s latest Gemini 3 model, Hassabis offered a rare glimpse into the timeline and technical challenges that still lie ahead, even as DeepMind maintains its position at the forefront of AI development.
“I think it’s sort of dead on track,” Hassabis said. “We are really happy with this progress. I think it’s an absolutely amazing model and is right on track of what I was expecting and the trajectory we’ve been on actually for the last couple of years, since the beginning of Gemini, which I think’s been the fastest progress of anybody in the industry.”
The DeepMind chief expressed confidence that this momentum would continue: “I think we are going to continue doing that trajectory and we expect that to continue. But on top of that, I still think there’ll be one or two more things that are required to really get the consistency across the board that you’d expect from a general intelligence.”
Hassabis identified specific areas requiring further breakthroughs: “Also improvements still on reasoning, on memory, and perhaps things like world model ideas that you also know we’re working on with Simmer and Genie. They will build on top of Gemini, but extend it in various ways and I think some of those ideas are going to be required as well to fully solve physical intelligence and things like that.”
Despite the challenges ahead, Hassabis maintained his optimism about Gemini 3’s reception: “I’m really happy with the progress of Gemini 3. I think people are going to be pretty pleasantly surprised, but it’s on track of what we were expecting the progress to be. And I think that means still five to 10 years with one or two more breakthroughs required.”
The timeline Hassabis outlined—five to 10 years until AGI—aligns with assessments from other leading AI researchers, though it represents a more measured stance than some of the more ambitious predictions circulating in Silicon Valley. His emphasis on needing “one or two more breakthroughs” in areas like reasoning, memory, and world models suggests that scaling up existing architectures alone won’t be sufficient. The reference to DeepMind’s work on Sima and Genie hints at the company’s exploration of AI systems that can better understand and simulate physical environments, a crucial capability for achieving true general intelligence. This measured optimism stands in contrast to more cautious voices in the industry who have questioned whether current approaches can deliver AGI at all, while also tempering the expectations of those predicting imminent breakthroughs. As competition intensifies between Google DeepMind, OpenAI, Anthropic, and others, Hassabis’s comments signal that the race to AGI remains very much active—but that the finish line, while visible, is still some distance away.