Google Gemini 3.5 Pro Model Delayed Due to Coding Issues

Google has delayed the launch of its Gemini 3.5 Pro AI model, a flagship product intended to compete with emerging rivals in the AI programming space, according to a Bloomberg report citing current and former employees.

The model, initially promised for a June rollout by Google CEO Sundar Pichai, has not been released despite Google’s efforts to address technical challenges. A late-June intervention involved updating the model’s training data to improve coding accuracy, but internal tests reportedly yielded unsatisfactory results, according to the Bloomberg report.

Internal structural issues have compounded the technical difficulties. Google’s fragmented corporate organization, with separate teams across Google Cloud, DeepMind, and the Android division working on parallel AI programming tools, has led to duplicated efforts and competition for computing resources. To address this, the company is consolidating these tools onto a centralized platform called Antigravity.

The delay comes as Google’s reliance on AI-generated code reaches an all-time high. Approximately 75% of new code deployed at the company is now AI-generated and manually approved by engineers, up from 50% last autumn, according to internal data. However, this shift has created friction among engineers who argue that critical systems should remain human-written to ensure quality.

Competitive pressures have intensified as rivals accelerate their own AI developments. Anthropic launched its Fable 5 architecture, while OpenAI released GPT-5.6 Sol, a model optimized for coding and cyber defense. Meanwhile, the Chinese lab Moonshot introduced Kimi K3, a 2.8-trillion-parameter open-source model. These advancements have put additional pressure on Google to deliver on its AI roadmap.

Enterprise clients using the lighter Gemini 3.5 Flash model, which was released on schedule in May, have provided mixed feedback. Figma praised the model’s balance of speed and quality, while the educational platform Platzi switched to Anthropic’s models, citing slower performance, structured data errors, and high pricing. Google has confirmed it is testing the Pro model with corporate partners and the US government, though the DeepMind website still lists the model as “coming soon.”

The delays highlight broader challenges for Google as it navigates the complexities of AI development. Internal bureaucracy and corporate resource wars have led to Gemini 3.5 Pro coding issues. Engineers involved in the project have expressed frustration, with some worrying the company is losing its edge to faster rivals.

As the AI landscape continues to evolve, Google’s ability to recover from these setbacks will be closely watched. The success of Gemini 3.5 Pro could determine the company’s position in the next phase of AI-driven software development, particularly as smaller startups and international labs push the boundaries of what is possible. For now, the delay underscores the high stakes of competing in a field where speed and performance are paramount.