Google was supposed to ship Gemini 3.5 Pro in June. It is now the third week of July and there is still no release date.
Google has pushed back the launch of its flagship Gemini 3.5 Pro model after it repeatedly missed internal coding benchmarks. For operators who build workflows around AI vendor roadmaps, the delay raises the question of how much to trust announced release dates from even the biggest labs.
Google was supposed to ship Gemini 3.5 Pro in June. It is now the third week of July and there is still no release date.
Bloomberg’s reporting says the holdup is coding performance. The model kept missing Google’s own internal benchmarks, so the team reset the training data and ran it again in late June. It came back short a second time. That is the part worth sitting with. This is not a startup working out of a garage. This is Google, with more compute and more researchers than almost anyone on earth, and the flagship model still is not clearing the bar the company set for itself.
Inside the company the delay has apparently caused real friction. Some researchers reportedly worry Google is losing the coding race to Anthropic and OpenAI outright. Sergey Brin has pushed to move faster. Other teams are holding a line that important code should stay human written, not generated.
I bring this up because it is a useful reset for anyone leaning on vendor roadmaps to plan their own work. If Google, with its resources, cannot hit a self imposed deadline on its own flagship model, the safe assumption for the rest of us is that any vendor’s next big release lands later than promised, and you build your plans around the model you have today, not the one on the slide.