When Google introduced Gemini 3.5 Flash at Google I/O in May 2026, it did so with a clear promise: Gemini 3.5 Pro, the full flagship version of the new model family, followed in June. June came and went. An internal target dated July 17th was leaked to the press and widely shared, generating great anticipation among developers.
July 17th also came and went. As of today, July 20th, there is still no publicly available Gemini 3.5 Pro and Google has not announced an official updated release date. Mashable published an article this week titled “Where Is Gemini 3.5 Pro? The Case of the Missing AI Model” that captures the sentiment well.
Otherwise, Google’s July 2026 Gemini updates were substantial. Gemini 3.5 Flash – the faster and more efficient companion model – has been available since I/O and performs well in developer adoption metrics. The company has shipped Gemini Spark, a proactive task-performing AI assistant, and Gemini 3.5 Live Translate, a real-time cross-lingual language tool.
But the absence of the Pro model – intended to compete directly with GPT-5.6 Sol and Claude Opus 4.8 at the border level – has created a visible gap in Google’s competitive position, just as the rest of the industry is shipping aggressively. For context on what Google delivered at I/O and what’s still to come, this overview covers the big picture: https://www.youtube.com/watch?v=Dj02ygCFfRU
Why it keeps slipping
9to5Google reported on July 16 that the Gemini 3.5 Pro delays were specifically due to encoding performance issues. Google teams identified benchmark results on coding and software development tasks that were below the internal bar they had set for a pro-tier model, and instead of shipping a model that underperformed its predecessor in a commercially critical category, they retired it for further work.
The complete rebuild — which reportedly triggered the July 17 internal target — was intended to address these gaps in coding capability without compromising the model’s strong performance in other areas such as reasoning, multimodal understanding, and handling long contexts.
The 2 million token Context Window, widely reported as the Gemini 3.5 Pro specification, has not been officially confirmed by Google, and Business Insider reported that pricing and API structure for the model also remain unconfirmed.
The combination of unconfirmed specifications, missed launch windows, and a rebuild process whose completion schedule Google refuses to publicly disclose has made Gemini 3.5 Pro one of the most talked-about delays in the current AI product cycle.
The competitive cost of waiting
The timing is more important than usual because the Gemini 3.5 Pro competitive landscape has changed significantly since the I/O announcement. GPT-5.6 Sol launched on July 9th. Kimi K3 launched on July 16th with 2.8 trillion parameters and borderline benchmark results. Grok 4.6 with 2 trillion parameters is still weeks away from completing training.
The frontier model landscape that Gemini 3.5 Pro would have entered in June 2026 is significantly more competitive in July 2026 and will be even more competitive by the time the model actually ships. Google’s engineering teams are clearly trying to get the model right rather than releasing it early – a defensible decision – but the window in which it will land as a clear competitive statement rather than a catch-up version is rapidly shrinking.
Quick links:
