Google Delays Gemini 3.5 Pro Launch to Rebuild AI Model After Programming Issues

Google has postponed the launch of Gemini 3.5 Pro, its most powerful artificial intelligence model, by several months beyond the originally planned date. Internal testing revealed that the model had not yet achieved the performance level targeted by the company, particularly in programming code generation and handling complex tasks.

According to a Bloomberg report, Google’s teams have rebuilt significant parts of Gemini 3.5 Pro after encountering issues with its ability to manage lengthy programming tasks and maintain accuracy during multi-step operations. The company has not announced a new official release date, with the model now delayed by several months compared to the initial schedule.

This delay occurs amid intense competition from OpenAI, Anthropic, and xAI, all of which have introduced new models aimed at professional use, programming, and autonomous agent operation. Google is also facing additional pressure following the departure of several prominent researchers from its teams to rival firms in recent months.

The news of the delay negatively impacted Alphabet's stock, Google's parent company, which fell approximately 4.4% during trading. Investors expressed concern over the widening gap between the company’s substantial investments in AI infrastructure and its ability to deliver a new model that can compete with the latest market offerings.

Google had previously released a limited trial version of Gemini 3.5 Pro to select corporate clients but continued refining its capabilities before a public rollout. Development efforts have focused on enhancing programming accuracy, reducing errors, and improving efficiency in executing extended tasks across multiple tools.