Google released three new artificial intelligence models on July 21, 2026: Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber. But the one many people were waiting for, the more powerful Gemini 3.5 Pro, was not included.

Think of Google's Gemini models like cars. The "Flash" versions are the efficient, everyday models built to be fast and affordable. The "Pro" versions are the high-end models with more power. Google had promised a new Pro model was coming soon, but it did not arrive in this release.

Google called Gemini 3.6 Flash its "workhorse model," saying it does better at coding, answering complex questions, and working with both text and images. It also uses less of what AI researchers call tokens, the small pieces of text the model reads and writes. Fewer tokens means lower costs for people using the model. Google said 3.6 Flash uses up to 17% fewer tokens than the previous version, according to its official blog (TechCrunch; Google Blog).

Gemini 3.5 Flash-Lite is designed to be the cheapest option. Gemini 3.5 Flash Cyber is specially trained to find and fix computer security problems. Google said it will only give the Cyber model to governments and trusted partners through a limited program, not to the general public (TechCrunch).

Google said these releases focus on making AI cheaper to run, faster to respond, and more reliable for businesses building AI-powered tools. Across the tech industry, companies are caring less about raw test scores and more about what it actually costs to use an AI model day to day.

The missing model, Gemini 3.5 Pro, is the bigger story. Google last updated the Pro tier in February 2026. In May, Google said 3.5 Pro was already being used inside the company and would be released the next month. That did not happen. Bloomberg reported on July 16, 2026 that Google was struggling to meet its own performance goals for the model (TechCrunch).

Logan Kilpatrick, Google DeepMind's product lead, said on July 21 that the company is testing 3.5 Pro with partners and hopes to "land soon." He also mentioned that the team has started its most ambitious training effort yet for Gemini 4, the next generation (TechCrunch).

The delay matters for paying customers. Google's AI Ultra plan costs $99.99 per month and includes access to Gemini 3.1 Pro with advanced research features (Reuters). Those subscribers have now gone five months without a Pro-tier update, even as the Flash line has been updated twice in that same period.

A separate Reuters report from July 20, 2026, citing The Information, said Google is building a new computer chip designed to run Gemini models more efficiently (Reuters). Combined with the efficiency gains in 3.6 Flash, this suggests Google is working to lower costs from two directions at once: better software and better hardware.

The Flash Cyber model raises its own questions. AI models trained specifically for security work are not entirely new, but limiting access to governments and trusted partners is a deliberate choice. Keeping it out of general circulation limits the chance of misuse, but it also means fewer independent experts can examine how well it works. Google did not say who counts as a "trusted partner" or what transparency will accompany the program.

The broader picture is that Google is running two speeds at once. The Flash models keep getting better quickly, with real cost savings. The Pro line has been stuck since February. The 17% token reduction in 3.6 Flash is the kind of improvement developers will notice directly on their bills.

In my view, if the Pro delay means Google is holding back a model that does not meet its own standards, that is probably the right decision. Releasing a flagship product that disappoints would hurt more than waiting. But Google promised this model in May, and it is now July. Kilpatrick's "land soon" is starting to sound less casual and more like a product that is genuinely stuck.

For now, the situation is simple. The Flash tier is improving fast. The Pro tier is frozen at a version from February. And somewhere in Google's systems, Gemini 4 is already being built.