During a conference call, Google disclosed that its latest, state-of-the-art model, Gemini4, has officially commenced training, with substantial resources earmarked for its development. The model is anticipated to make its debut in November or December of this year. Earlier, the planned release of the Gemini 3.5 Pro in June did not materialize, prompting discussions about the pace of Google's model advancements. In reaction, Pichai emphasized that the competition in the cutting-edge model sector is intense, yet Google continues to maintain a competitive edge in various areas. At the same time, he conceded that Gemini has certain limitations in critical domains like programming and intelligent agents. He sees Gemini4 as pivotal in bridging these gaps and is pleased with the progress of its training. The timely release of Gemini4 and its ability to fulfill expectations will undoubtedly be a significant focal point in the AI large model arena this year.