Google’s been on a tear with Gemini releases lately, but its next big one isn’t showing up on time. The Gemini 3.5 Pro delay is real, according to Bloomberg, citing ten current and former employees.
Coding is the sticking point. Anthropic and OpenAI have both shipped models that outperform Google’s current lineup at writing code. Gemini 3.5 Pro was supposed to close that gap. Google reportedly refreshed the model’s training data last month specifically to improve coding, but the results fell short internally. Sundar Pichai said back in May that 3.5 Pro was already running internally and would roll out the following month. That month came and went. So far, Gemini 3.5 Flash is the only model in the 3.5 lineup that’s actually shipped.
Why It’s Taking So Long
Part of the problem is Google’s own size. Cloud, DeepMind, and the Android team are all reportedly building separate AI coding tools. Getting a model through that many stakeholders slows things down. Engineers are also said to be competing with each other just to get computing power for their own testing. The delay has apparently pushed some frustrated staffers out the door toward Anthropic and other labs. Alphabet shares dropped a few percent on the news.
There’s a wrinkle worth watching outside Google too. Apple’s newest Foundation Models were built by distilling a custom Gemini model as part of a multi-year deal. Phandroid has previously covered how a stronger Gemini release benefits Apple’s Siri work too. Whether the Gemini 3.5 Pro delay trickles down to Apple’s roadmap isn’t clear, and nothing confirms it will.
Still, if Google’s frontier model work keeps slipping, it’s worth wondering whether Apple’s future AI plans feel it too. Apple spent years shaking off a similar dependency with Intel, back when Mac upgrades were effectively gated by Intel’s own chip schedule. Leaning on Gemini for training puts a version of that same risk back on the table, even if it’s a much smaller piece of Apple’s roadmap this time.
Google, for its part, says it’s shipping quickly across a wide range of models while keeping them cost-effective. Whether that holds up once the Gemini 3.5 Pro delay finally ends is another question entirely.
