Google wanted to ship Gemini 3.5 Pro this year. That is not happening now.
The company has pushed the launch back by months while it tries to make the model better at writing code. That is the part where rivals have pulled ahead, and Google knows it. In late 2022, following ChatGPT's debut, Google issued a company-wide "code red" alert to accelerate its AI development. Competing companies such as OpenAI and Anthropic have since unveiled AI systems that surpass what Google currently has available.
According to a Google spokesperson, the company is still evaluating Gemini 3.5 Pro, a newer Flash variant, and several other models. A Google spokesperson said, "We're shipping quickly across a wide range of models while keeping them highly cost-effective for customers."
Get the market news that matters in a five-minute read with Market Briefs, our free daily newsletter
The delay is not just about the technology. It is also about the people and the process inside Google.
Current and former employees say there are competing factions across product teams, disagreements among engineers about what role AI should play in coding, and plain old internal bureaucracy slowing things down. Google co-founder Sergey Brin has pushed for faster AI coding efforts. To address the disarray, Google is consolidating its AI coding initiatives under Chief AI Architect Koray Kavukcuoglu and establishing a specialized coding team within DeepMind, headed by research engineer Sebastian Borgeaud.
As businesses increasingly adopt AI-assisted development tools, coding skill has emerged as a key measure of AI model performance. Google's internal use of AI to generate 75% of its code demonstrates the potential, but externally, Gemini must achieve competitive benchmark scores to retain customers. The departure of key researchers to labs like Anthropic has depleted the expertise needed to close this gap, and organizational inefficiencies further slow iteration. This exodus of talent and internal friction has made it difficult for Google to iterate quickly, compounding the technical challenges.
But for the Gemini model itself to compete with the best from OpenAI and others, the coding skills need to be top-tier. Some customers are already shifting to competitor models.
Coding benchmarks have become a critical differentiator in the AI race. Models like GPT-4o and Claude have set high standards, and enterprises increasingly demand reliable code generation for software development workflows. The fact that Google employs AI to produce three-quarters of its own code illustrates the opportunity, though it must demonstrate comparable external performance to retain enterprise customers.
The stakes are high for Google. The company's cloud division, a key growth driver, relies on competitive AI models to attract enterprise clients. If Gemini cannot match the coding performance of OpenAI's GPT-4o or Anthropic's Claude, Google risks losing ground in the lucrative AI-as-a-service market.
Join Market Briefs, our free daily newsletter, for a quick daily rundown of the markets
