Google DeepMind rolled out three fresh additions to its Gemini lineup this week: Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber. All three lean into the same set of priorities — speed, efficiency, and dependability — aimed squarely at customers building AI agents at scale. What's missing from the drop, though, is the piece a lot of people were actually waiting on: an update to Gemini Pro.

What Google Actually Released

The headliner is Gemini 3.6 Flash, which Google frames as its everyday "workhorse model." It's meant to handle the bread-and-butter work — coding, knowledge tasks, and multimodal jobs — with better results than before. The efficiency angle is the real pitch here: the model trims token usage by as much as 17%, which brings the running cost down below its predecessor, 3.5 Flash. Cheaper and more capable at the same time is a hard combination to argue with.

Sitting below that is Gemini 3.5 Flash-Lite, positioned as the most budget-friendly option in the class. If you're running high volume and every fraction of a cent matters, that's the one Google is pointing you toward.

A Model Built Specifically for Security Work

The third release is the interesting one. Gemini 3.5 Flash Cyber is a specialized model, fine-tuned to find and fix cybersecurity vulnerabilities, and Google says it does that at a reasonable price. But you can't just sign up for it. Access is locked to governments and trusted partners through a limited pilot program — a fairly deliberate choice for a tool built around probing and patching security holes.

Taken together, the three models tell a consistent story. Google isn't chasing raw reasoning horsepower with this batch. It's tuning for efficiency, latency, and reliability — the qualities that matter most when you're running AI agents in production rather than testing them in a lab.

The Gap Everyone Noticed: No Gemini 3.5 Pro

Here's the thing about this launch — what Google left out says as much as what it shipped. There's still no refreshed Gemini Pro. The flagship line last got an update back in February, and Pro is generally where Google puts its highest-capability work: complex reasoning and heavier coding tasks. Flash models, by contrast, are the ones built for lower cost and faster responses in real-world applications. So shipping a wave of Flash variants while Pro sits untouched leaves a noticeable hole at the top of the stack.

Google hasn't been quiet about Pro being on the way. Back in May, alongside the 3.5 Flash release, the company said the Pro version was already running internally and that a public rollout was coming the following month. That timeline has clearly slipped. Bloomberg reported last week that Google has hit internal delays with 3.5 Pro, struggling to clear its own performance targets before pushing it out.

Why the Timing Stings

The delay lands during a stretch where Google's competitors haven't slowed down at all. Since that February Pro update, OpenAI put out GPT-5.5 and started rolling out GPT-5.6. Anthropic, meanwhile, launched Claude Opus 4.8 and Claude Sonnet 5, and widened access to its frontier Fable 5 model. That's a lot of ground covered by rival labs in a short window — which makes a quiet Pro line stand out more than it otherwise would.

What Google Is Saying About What Comes Next

There's reassurance coming from inside the team. Google DeepMind product lead Logan Kilpatrick said this week that 3.5 Pro is currently in testing with partners and that the company hopes to land it soon. He also mentioned something with a longer horizon: the team has kicked off its most ambitious pre-training run yet for Gemini 4.

So the picture is a little split. On one hand, the current release is all about practical, cost-conscious models for production use. On the other, the marquee reasoning update is stuck in the wings, and Google is already looking past it toward a much bigger swing with Gemini 4.