Google has pushed back the release of Gemini 3.5 Pro by several months, a setback that has raised pressure inside the company as Anthropic and OpenAI keep rolling out stronger AI products. According to Bloomberg, Google had expected to launch the flagship model after its developer conference in May 2026, but the timeline has slipped sharply.
Launch falls behind as coding performance remains a weak spot
Bloomberg said the main obstacle is that Gemini 3.5 Pro has not yet surpassed rivals in code generation and code analysis. Google updated its training data at the end of last month in an attempt to improve the model, but the early results fell short of expectations.
That leaves the company under real pressure. Anthropic and OpenAI have continued to release better-performing products, while newer models from OpenAI and Meta have recently taken the lead in programming-related tasks.
Responding to outside scrutiny, a Google spokesperson said the company is actively testing the new model with partners and is in communication with the US government over its safety architecture.
Internal structure has slowed product development
The report also points to organizational issues inside Google. Google Cloud, the DeepMind research lab and the Android team are all developing their own AI coding tools, which has led to duplicated work, internal competition and poor coordination of research resources.
Google’s earlier effort to guard against intellectual property leaks also limited engineers’ ability to use Gemini internally. That, in turn, reduced the speed of model iteration.
Talent losses and compute constraints add to the strain
The slower progress has been accompanied by departures among research staff. Some top researchers have moved to competitors including Anthropic. Engineers inside Google have also run into compute bottlenecks when developing and running AI code because of internal competition for computing power.
Those structural problems have weighed on team morale and made it harder for Google to sustain its long-standing AI leadership strategy.
Google turns to Antigravity to unify its tooling
To address the disorder in development, Google has started an initiative called Antigravity. Its aim is to bring scattered coding tools together under a standardized framework for safety and memory protocols.
Google also said that 75% of its internal production code is now generated by AI and then reviewed by humans. The company added that its models still hold distinct advantages in multimodal capabilities and search-based querying, and said it plans to retain enterprise customers by improving cost efficiency.