Gemini 3.5 Pro Delayed: Google Faces Internal Challenges in Development

Google has reportedly postponed the launch of Gemini 3.5 Pro, the AI model considered the most powerful in its family, because the system has not yet met the internal targets set by the company. According to several sources familiar with the situation, the delay is primarily related to the need to improve coding and writing capabilities.

This decision has generated discontent among various engineers, researchers, and project leaders in Google's AI division, who are concerned that the group may lose ground to competitors like Anthropic and OpenAI, which in recent months have released models that perform better in certain areas, especially programming.

One of the main issues seems to be tied to internal organizational complexity. Google involves numerous teams and divisions in preparing the models, aiming to integrate AI into many products, including Search, Maps, and YouTube. However, this approach can slow down decision-making processes and create difficulties in coordinating resources.

Google is trying to enhance Gemini 3.5 Pro across various fields. The company recently updated the data used to train Gemini with the aim of boosting software development capabilities, but initial results have reportedly not met expectations. Nevertheless, Google has stated that it continues to work rapidly on different models while keeping an eye on costs and collaborating with partners and institutions.

Competition in the AI sector has intensified following the success of ChatGPT in 2022. Google had responded by declaring an internal emergency, but today the race for artificial intelligence has become a permanent condition for the company.

Within the group, there are also different approaches: some divisions such as Google Cloud, DeepMind, and Android are developing tools dedicated to AI-assisted programming, while some engineers have expressed doubts about the heavy use of artificial intelligence in code production. However, Google claims to have made significant progress, asserting that a large portion of the code produced internally is now generated with the help of AI and subsequently verified by developers.