According to Sundar Pichai, CEO of Google, during the second-quarter earnings call in 2026, the next-generation cutting-edge model Gemini4 has officially entered training, and Google has invested "a significant amount of resources" into it. Based on Google's past release schedule, the outside world generally expects Gemini4 to be officially unveiled in November or December of this year.
Admitting Lag in Programming and Agents, Counting on Gemini4 to Catch Up
Notably, the previous Gemini 3.5 Pro was not released as planned in June, which has raised questions about Google's model iteration rhythm. In response to analysts' sharp questions about whether "Google can maintain its industry leadership," Pichai admitted that the cutting-edge model competition is extremely dynamic, with a rapidly changing competitive landscape. However, he also emphasized that Google has previously launched multiple industry-leading models and continues to maintain advantages in multiple dimensions.
Pichai did not avoid Google's current shortcomings—he admitted that Gemini still has gaps in key areas such as programming and agents, and the team is highly focused on these directions. He positioned Gemini4 as a critical step to narrow the gap, expressing his "very positive feelings" about the current progress and believing that the user experience will provide a satisfactory answer. Considering the continued leadership of Anthropic's Claude series in the programming field and OpenAI's accelerated iteration, whether Gemini4 can fulfill Pichai's promises as the end of the year approaches will be the biggest highlight of the final showdown in AI large models in 2026.
