Google Drops Three New Gemini Models, But the One Everybody Wanted Is Still Missing
Google DeepMind released cheaper, faster AI models on Tuesday, including one built just for cybersecurity. Its flagship 'Pro' model, however, is still delayed.
Key points
- Google DeepMind released three new Gemini models on Tuesday: Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber.
- Gemini 3.6 Flash cuts the amount of data it processes by up to 17%, making it cheaper to run than its predecessor.
- Gemini 3.5 Flash Cyber, a model trained specifically to find and fix security weaknesses in software, will be available only to governments and vetted partners.
- Google's flagship high-capability model, Gemini Pro, remains delayed after Bloomberg reported internal teams are struggling to hit performance targets.
- Google DeepMind's product lead confirmed the company has begun its biggest-ever training run for the next-generation Gemini 4.
Google DeepMind shipped three new artificial intelligence models on Tuesday, and they are cheaper, faster, and more specialised than what came before. What they are not is the model Google's customers have been waiting for.
The headliner of the trio is Gemini 3.6 Flash, which Google describes as its everyday workhorse. It handles coding, research tasks, and understanding images or documents, and it does so while using up to 17% fewer tokens. A token is roughly a word-fragment: the more tokens a model chews through, the more it costs to run. Fewer tokens means lower bills for businesses building apps on top of it.
Next is Gemini 3.5 Flash-Lite, billed as the cheapest option in the range. It targets developers who need fast, affordable answers at high volume, think customer-service bots or document-sorting tools.
The third model is the most unusual. Gemini 3.5 Flash Cyber is a version fine-tuned, meaning trained further on specialist material, to find and fix security vulnerabilities, the weak spots in software that hackers can exploit. Google will not sell it publicly. It is available only to governments and trusted partners through a limited pilot programme.
Why does the missing Pro model matter?
It matters because competitors are not standing still. Google last updated its flagship Gemini Pro model, the most capable version built for complex reasoning and hard coding problems, back in February. Since then, as first reported by TechCrunch AI, OpenAI has released GPT-5.5 and started rolling out GPT-5.6, while Anthropic has shipped Claude Sonnet 5 and expanded access to its own frontier model.
Google had hinted that a Pro update was coming. In May, the company said the new Pro version was already being used internally and would ship "next month." It did not. Bloomberg later reported that internal teams were struggling to hit their own performance targets before launch.
On Tuesday, Google DeepMind product lead Logan Kilpatrick said the team is testing Gemini 3.5 Pro with partners and hopes to "land soon." He also confirmed that Google has started its most ambitious training run yet for Gemini 4, the next major generation of the model family.
For ordinary users, none of Tuesday's releases change anything immediately visible. These Flash models power the back end of business tools, not the Gemini chatbot most people interact with. The practical upshot: apps built on Google's AI may get a little snappier and cheaper to run, but the big capability leap is still waiting in the wings.
