A late-June training-data fix disappointed, no new date is set, and Alphabet shares fell 4% on the report.
Google's Gemini 3.5 Pro, the company's most powerful model, has slipped months past the June launch target it set in May.
At its I/O developer conference in May, Google unveiled the Flash version of the model, and CEO Sundar Pichai said the Pro edition would follow the next month. That June deadline came and went with no fresh release date from the company.
The holdup centers on coding. In late June, Google refreshed the training data behind Gemini in the hope of sharpening those skills, but the results fell short of expectations.
The slip has frustrated staff across engineering, AI research and management, Bloomberg reported. Ten current and former employees told the news organization they worry Google risks losing its competitive edge as rivals Anthropic and OpenAI ship models that outperform Gemini.
The internal picture is contradictory, Bloomberg reported: engineers are now expected to write code with the help of AI, yet they often run into capacity limits as teams compete internally for computing power.
Google pushes back. A spokesperson said the company is trialing version 3.5 Pro alongside an upgraded Flash and other models with partners, and is shipping models quickly across a broad lineup while keeping them affordable for customers. Google has also said that, as of April, 75% of its new code is generated by AI and then approved by engineers, up from 50% the previous fall — a figure that sits against the constraints its own engineers describe.
For now, no successor date has been named, leaving Gemini 3.1 Pro, which dates to February, as Google's current flagship.