# Google released three Gemini models in one day while its flagship is still stuck in testing > Source: > Published: 2026-07-22 00:34:26+00:00 *Google dropped Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber on July 21, 2026, flooding its mid-tier lineup while Gemini 3.5 Pro, originally due in June, has now missed three consecutive launch deadlines.* Three new models in a single day is a bold move. It is also, if you read the subtext, a defensive one. Google DeepMind announced Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber simultaneously on Tuesday, cutting prices and chasing efficiency gains across the board, all while the flagship model the company promised at Google I/O in May still has no public launch date. Gemini 3.6 Flash is the headline release. It's the successor to 3.5 Flash, and according to Google, it generates roughly 17% fewer output tokens on multi-step tasks by tightening its reasoning steps. That efficiency gain translates directly to cost: output pricing drops from $9 to $7.50 per million tokens, with input sitting at $1.50 per million. Google is calling it its new workhorse model, improved across coding, knowledge work, and multimodal tasks. Both 3.6 Flash and 3.5 Flash-Lite are available now to developers through the Gemini API via Google AI Studio and Android Studio. Flash-Lite is the cheapest option in the lineup and is slated to roll into Google Search. It's aimed squarely at high-volume, cost-sensitive applications where raw power matters less than throughput and price. These aren't the models that win benchmark headlines. They're the ones that win contracts. Gemini 3.5 Flash Cyber is a fine-tuned variant built specifically to find, validate, and patch software vulnerabilities. Google tested it against the V8 JavaScript engine, where it turned up 55 unique confirmed vulnerabilities, compared with 47 for the standard 3.5 Flash and 36 for Anthropic's Claude Opus 4.6, according to Google DeepMind's own benchmarks. Whether you take those numbers at face value or not, they make the competitive framing plain. This is a direct push into enterprise security tooling, a space where both Anthropic and OpenAI have been actively positioning. Google is citing the dual-use risk of the technology, so access to the Cyber model is restricted to governments and trusted partners through a limited pilot, at least for now. The timing matters. Anthropic has been building its Claude-for-security story for months. OpenAI has its own security-oriented offerings. For Google to enter with a fine-tuned model, benchmark data in hand, and distribution restrictions that signal serious intent rather than casual exploration, this isn't a prototype. It's a product with a target market. ## The flagship problem Frankly, the more revealing story is what didn't ship. Gemini 3.5 Pro was announced at Google I/O in May and originally expected in June. It has now missed three consecutive launch deadlines, as reported by 9to5Google and AndroidHeadlines. The issues, per multiple reports, center on coding performance. Google updated the model's training data in late June to address the gap, but the results reportedly fell short of benchmarks set by GPT-5.6. Google's current public position is that it's testing 3.5 Pro and other models with partners, with no release date on offer. That's a problem. The Flash tier handles volume and cost efficiency, but it's the Pro model that signals whether Google can compete with frontier reasoning. OpenAI and Anthropic have each shipped significant reasoning updates in the past few months, and Google's flagship is still in the lab. Three new Flash variants don't paper over that gap, though they do give developers real options while they wait. There is one genuinely good piece of news buried in Tuesday's announcement: Google confirmed it has started pre-training on Gemini 4. That's early enough to be mostly a signal rather than a roadmap, but it does suggest the company isn't content to camp indefinitely in the 3.5 generation. The picture that emerges is a company executing well at the mid-tier while fighting to stay competitive at the top. Three models in one day, with genuine price cuts and a credible security play, shows Google still knows how to ship. The question is whether it can close the Pro-tier gap before that story becomes the only one anyone wants to tell. **Also read:** [Intel cuts jobs in its fastest-growing division two days before Q2 earnings](https://startupfortune.com/intel-cuts-jobs-in-its-fastest-growing-division-two-days-before-q2-earnings/) • [Supermicro books $60 billion in orders in a single quarter as AI server backlog hits record high](https://startupfortune.com/supermicro-books-60-billion-in-orders-in-a-single-quarter-as-ai-server-backlog-hits-record-high/) • [OpenAI adds two bank CEOs to its board as IPO signals grow stronger](https://startupfortune.com/openai-adds-two-bank-ceos-to-its-board-as-ipo-signals-grow-stronger/)