Google has postponed the launch of its Gemini 3.5 Pro AI model, moving the release date from June to July 2026, as it focuses on improving the model's coding capabilities. The delay comes amid rising competitive pressure from companies like OpenAI and Anthropic, which have recently released superior models.
Employees familiar with the situation express frustration about the setback, fearing that the tech giant might fall behind its competitors. Google had initially planned to showcase Gemini 3.5 Pro at its developer conference in May, but disappointing results from recent updates to the model's training data necessitated additional work.
Following this news, Alphabet's stock experienced a decline of up to 3.2% on Thursday. The company is currently testing Gemini 3.5 Pro alongside an upgraded Flash model and other systems with various partners. Discussions with the US government regarding the model's capabilities and standards are also ongoing.
Internal Challenges and Competition
As OpenAI and Meta release models that excel in coding tasks, researchers within Google recognize access to Search data as Gemini's key advantage. However, competitors are outperforming Google in overall model capabilities. Compounding these challenges is Google's size, which has led to fragmentation among its various teams, including Google Cloud, DeepMind, and Android, all developing separate AI coding tools. This has resulted in overlapping efforts and competition for resources.
Executives like Sergey Brin are urging a faster pace in AI coding innovations, but internal divisions are proving to be a significant hurdle. Some engineers initially resisted AI-generated code, asserting the need for human-written software to meet Google's standards. Concerns about proprietary information entering the training data of Gemini also restricted earlier use of the AI for analyzing internal code.
However, these restrictions have recently been relaxed, and Google now expects engineers to utilize AI for coding tasks. The company reports that 75% of its code is now generated by AI, subject to employee review before being put into production. To streamline its efforts, Google has consolidated several coding products under the Google Antigravity umbrella, which provides essential infrastructure for AI systems interacting with applications.
Despite these initiatives, employees still encounter capacity limits when using AI tools, as different teams compete for computing resources. This ongoing frustration has led some researchers to leave Google for positions at competing firms such as Anthropic. While customers report mixed experiences with Gemini 3.5 Flash, they await the arrival of the anticipated Pro version.
This article is for informational purposes only and should not be considered financial advice.
