Google Gemini Launch Delayed as Tech Falls Short of Internal Goals

Google has quietly pushed back the release of Gemini 3.5 Pro, and the holdup comes down to something pretty specific: the model isn't coding as well as leadership wants it to. That's not a small gap to close. Coding benchmarks have become the industry's favorite yardstick, the thing every lab points to when they want to prove they're ahead of OpenAI or Anthropic, and Google apparently isn't satisfied with where 3.5 Pro currently lands.

The delay is now stretching into months, according to people familiar with the effort, and it's stirring up frustration inside the company. Engineers and researchers who worked on earlier Gemini releases are reportedly uneasy watching the schedule slip, and managers are worried about what a slower cadence means competitively. Google spent much of this year building momentum with Gemini 2.5 and its various Pro and Flash variants, closing a gap that once looked embarrassingly wide. A stumble now, even a modest one, risks handing that narrative back to rivals.

What makes this notable is that Google isn't just tweaking a minor feature or waiting on safety review. The company is trying to push a genuine capability jump in coding, an area where developers vote with their feet and switch tools the moment a rival model writes cleaner, more reliable code. Missing that bar internally suggests Google's own team believes shipping a mediocre coding upgrade would do more harm than staying quiet a little longer.

There's also a broader signal here about how tight the margins have gotten between the top AI labs. A few months of delay didn't used to matter much when foundation model releases came a year or more apart. Now, in a cycle where competitors ship meaningful updates every few weeks, even a modest slip can look like falling behind, whether or not the underlying model actually is.