Skip to main content

Google CEO confirms Gemini 4 is training, flagship AI model expected by year-end

Google CEO Sundar Pichai dropped a major update during the company's second-quarter earnings call in 2026: the next-generation flagship AI model, Gemini 4, has officially begun training. Pichai confirmed that Google has poured "a significant amount of resources" into the project, signaling its importance to the company's AI strategy.

Based on Google's historical release patterns, industry watchers expect Gemini 4 to be unveiled in November or December of this year. The announcement comes amid heightened scrutiny of Google's AI roadmap, especially after the previous model, Gemini 3.5 Pro, failed to launch as planned in June.

Acknowledging the competition

During the call, Pichai faced tough questions from analysts about whether Google can maintain its leadership in the rapidly evolving AI landscape. He didn't shy away from the challenges, admitting that the competitive field is "extremely dynamic" and that the landscape is changing quickly. However, he also pointed out that Google has previously launched multiple industry-leading models and continues to hold advantages in several areas.

More notably, Pichai openly acknowledged specific shortcomings in Gemini's current capabilities. He admitted that the model still lags behind in key areas like programming and agent functionality—two domains where competitors like Anthropic's Claude series and OpenAI's GPT models have made significant strides. The Gemini team, he said, is "highly focused" on closing these gaps.

Gemini 4 as the answer

Pichai positioned Gemini 4 as a critical step in narrowing the gap. He expressed "very positive feelings" about the current progress and believes that the user experience will ultimately provide a satisfactory answer to skeptics. "We're investing heavily, and I'm optimistic about what we're seeing," he said.

The stakes are high. Anthropic's Claude series has continued to lead in programming benchmarks, while OpenAI has accelerated its iteration cycle with new models. As the year-end approaches, all eyes will be on whether Gemini 4 can deliver on Pichai's promises and reclaim Google's position at the forefront of AI innovation.

Image

Key Points:

  • Gemini 4 has officially entered training with significant investment from Google.
  • Expected launch window: November or December 2026.
  • Pichai acknowledged gaps in programming and agent capabilities compared to competitors.
  • The model is seen as a critical step to catch up with Anthropic and OpenAI.
  • The year-end release will be a major highlight in the AI large model race.