Google Ships a Trio of Gemini Flash Models While 3.5 Pro Stays MIA and Gemini 4 Training Begins

Google Ships a Trio of Gemini Flash Models While 3.5 Pro Stays MIA and Gemini 4 Training Begins

Google launched three new Gemini family models on July 21, 2026: Gemini 3.6 Flash, Gemini 3.5 Flash-Lite and Gemini 3.5 Flash Cyber, according to a TechCrunch report echoed by Brazilian outlet Boa Informação. The update came with an admission unusual for a technology company: the most anticipated model in the batch, Gemini 3.5 Pro, still has no broad launch date, after successive delays.

Three New Models, Three Different Jobs

Google describes Gemini 3.6 Flash as the family's workhorse, with gains in coding, knowledge tasks and multimodal use, plus up to a 17% cut in token consumption compared with the previous generation. Gemini 3.5 Flash-Lite aims at the opposite end: the cheapest option in the lineup, built for high volume and low latency. Gemini 3.5 Flash Cyber, meanwhile, is a variant tuned specifically to find and fix cybersecurity vulnerabilities, and Google chose to restrict its access to governments and trusted partners, with no broad public release.

The Model That Promised June and Still Hasn't Arrived

Gemini 3.5 Pro was announced at Google I/O in May 2026, promised for broad rollout within a month. That deadline slipped in June, was informally reset to July 17 according to reporting from TechTimes and BigGo Finance, and also failed to materialize. As of this article, the model remains restricted to testing with select partners, with no public date for general availability.

Meanwhile, Google Is Already Eyeing Gemini 4

In the same July 21 announcement, Google revealed it has begun what it calls its most ambitious pretraining run yet, aimed at Gemini 4, according to a roundup from BuildFastWithAI. The move suggests Google decided not to wait for 3.5 Pro to mature before moving forward on its roadmap, a choice that exposes the competitive pressure of the moment: OpenAI, Anthropic, xAI and Moonshot AI have all shipped new frontier models in just the past two weeks.

Why It Matters to Anyone Automating Customer Service and Marketing

For agencies and SMBs running integrations on top of Google's API, the practical lesson repeats itself: the model actually in production remains the Flash generation, not the promised Pro. Anyone planning architecture, budget or a commercial offer around an expectation of a near term Gemini 3.5 Pro risks delaying their own delivery to clients. It is worth treating the new Flash models, already available, as the immediate operational baseline, and reserving any plan that depends on the Pro tier for whenever it actually leaves restricted preview.

What's Clear

Google keeps shipping models at a fast pace, just not in the category it needs to close the gap with frontier rivals. While 3.5 Pro piles up delay after delay, the Gemini 4 training announcement signals the company would rather skip ahead than stay hostage to a model that isn't closing the gap. For whoever decides which AI vendor to use, the yardstick keeps being what's actually available in production, not what shows up in a keynote.