Google Delays Flagship Gemini 3.5 Pro Model as Coding Gaps Widen
Alphabet's Gemini 3.5 Pro AI model is months behind schedule due to coding performance shortfalls, as rivals OpenAI and Meta pull ahead in code generation…
Google is pushing back the release of its most advanced AI model after discovering it can't match rivals in a critical task: writing software code.
Alphabet shares dropped 4 percent Thursday following Bloomberg's report that the Gemini 3.5 Pro model is months behind schedule. The company had announced the model in May during its annual Google I/O developer conference, saying it was being tested internally and would roll out to users the following month. That timeline has now slipped significantly.
The hold-up centers on coding. According to sources familiar with the matter, the model's ability to generate software code fell short of Google's internal expectations—a particularly damaging gap at a moment when OpenAI and Meta have recently released competing models that outpace Google's current offerings in this area.
Code generation has become one of the biggest use cases for AI vendors. Software developers rely on these tools to write, debug, and optimize code faster, making performance here a direct measure of a model's usefulness in the enterprise market.
An Alphabet spokesperson told CNBC the company remains focused on shipping models responsibly. "We're currently testing 3.5 Pro, an upgraded Flash model, and other models with partners, and we're productively engaged with the U.S. government," the statement said. The company emphasized it is "shipping quickly across a wide range of models while keeping them highly cost-effective for customers."
The delay underscores mounting pressure on Google as it battles to retain leadership in the AI race. Gemini, which launched in December 2023 and replaced Google's earlier Bard chatbot, has expanded rapidly across the company's ecosystem. The latest version, Gemini 3.1 Pro, arrived in February. Earlier this summer, Google rolled out a suite of upgrades to its Gemini apps, expanding their reach across more Google services and making them harder to avoid for anyone using the company's products.
But growth alone hasn't solved the core challenge: building models that consistently outperform OpenAI's ChatGPT family and newer entrants like Meta's Llama. The Gemini architecture, trained natively on multiple data types, can process and generate text, code, images, audio, and video simultaneously. Google distributes the technology across different tiers—from on-device "Nano" versions to high-compute "Pro" and "Ultra" models—each targeting different use cases and price points.
The broader product strategy has also shifted. Google recently revamped how it meters Gemini usage across its Free, Plus, Pro, and Ultra tiers. Rather than counting requests, the company now measures usage by the computing power each prompt requires. This gives Google tighter control over costs but leaves users less certain about when they'll hit limits. According to the company's own support documentation, "access is subject to change or may be limited based on testing, experimentation or availability."
For developers and enterprises, the 3.5 Pro delay signals that Google is taking performance seriously—but it also means waiting longer for a model pitched as a step forward. The company's iOS app received an update just days ago on July 14, bringing Gemini to more Apple devices, suggesting Google is still pushing the broader platform even as the flagship model remains under wraps.
