Skip to main content

Google's Gemini 4 Enters Post-Training, Could Arrive Before Year-End

Google's Gemini 4: The Next Big Thing, and It's Coming Soon

At The Information's AI Agenda Live event on September 23, Demis Hassabis, the head of Google DeepMind, dropped a bombshell: Google's next flagship model, Gemini 4, is already in its early post-training phase. In plain English, that means the base model is now being fine-tuned and made more reliable before it hits the public.

Hassabis didn't hold back his enthusiasm. He hopes Gemini 4 will launch "well before the end of the year" and said the team is "rapidly delivering early results" from post-training. "Because we saw the results and were very excited," he explained. After that, Google plans to keep iterating at a fast clip.

Why Skip Gemini 3.5 Pro?

You might remember that back in May, Google CEO Sundar Pichai promised a major update to Gemini 3.5 Pro would land in June. It never showed up. Hassabis shed some light on that: after Gemini 3 and 3.1, Google decided to "take a step back" and focus on smaller, cheaper Flash models instead. "For us, the most important thing at that time was to accelerate learning as much as possible," he said.

So, was 3.5 Pro cancelled? Hassabis didn't say outright, but he made one thing clear: all eyes are on Gemini 4.

Already in Use, Safety Checks Underway

Gemini 4 isn't just sitting in a lab. Google engineers have already been using it internally to run the Antigravity AI programming tool. Meanwhile, the model is going through safety testing and protective mechanism development.

When asked if he's worried about falling behind competitors like OpenAI or Anthropic, Hassabis didn't flinch. "I have 100% confidence in the team," he said. "Our team is incredibly strong, and I have the greatest trust in them; we will always be at the forefront of technology."

The AGI Question: A Distraction?

Unlike some rivals who love to debate whether we've achieved AGI (artificial general intelligence), Hassabis sidestepped the topic. "Discussing whether AGI has been achieved is not the right direction," he said. "The real issue is whether we can build trustworthy agents."

That's a subtle but important shift. While others chase the AGI label, Google seems more focused on making AI that people can actually rely on.

Key Points

  • Gemini 4 is in post-training and could launch well before the end of the year, according to Demis Hassabis.
  • Google skipped Gemini 3.5 Pro to focus on Flash models and accelerate learning; the company is now all-in on Gemini 4.
  • The new model is already being used internally for coding tasks and is undergoing safety testing.
  • Hassabis expressed full confidence in his team despite fierce competition.
  • He believes the AGI debate is a distraction; the real goal is building trustworthy AI agents.