Gemini 3.5 Pro Misses Its Target Date Again

July 17 was supposed to be the day. For weeks, the widely repeated word was that Google would finally ship Gemini 3.5 Pro, the flagship model it tore down and rebuilt from scratch. The day arrived, and the model did not. As of this afternoon there is no model card, no pricing page, and no official general-availability announcement from Google, whose own documentation still says only that Pro is "coming soon."

This is the third target the model has slipped past. Google first pointed at June, then at a July window, and community leaks converged on July 17 to line up with the opening of the World AI Conference in Shanghai. None of it came from Google directly. As we noted when we first covered the delay and the full rebuild, every specific claim about dates, context windows, and benchmarks has come from third-party reporting rather than the company.

Why the rebuild dragged

The reason Google gave testers was quality. Rather than iterate on the original base model, the team scrapped it and started over, chasing gaps that early users flagged in mathematical reasoning, complex image generation, and the handling of intricate visual scenes. The logic is competitive. A model that lands well behind Anthropic's Fable 5 or OpenAI's newly public GPT-5.6 in head-to-head tests would be worse for Google than one that arrives late. Reporting this week suggests the rebuilt version is still throwing off hallucinations and inconsistent output that keep it below the bar the team set.

According to a July 16 report from Tech Times, Google is now considering a stopgap release, some interim version to put in developers' hands while the full model is finished. The company has not confirmed that plan. Prediction markets tracking the launch have moved their money to later in the summer, with late July and early August now the favored windows.

The cost of arriving last

The delay is awkward because the rest of the field has not waited. OpenAI shipped GPT-5.6, xAI put out Grok 4.5, and Chinese labs have flooded the open-weight market. Google remains one of the few frontier labs whose current-generation flagship is stuck in preview while its rivals sell theirs.

None of this means the model is in trouble. A lab choosing to hold a release rather than ship something it considers half-baked is defensible, arguably the responsible call. But the gap between the hype cycle and the shipping date has become the story, and every missed date makes the eventual launch a harder act to live up to. We will update this piece when Google puts a real model in front of the public.