Google DeepMind shipped three new Gemini models focused on speed, efficiency, and cybersecurity. The headliner, Gemini 3.6 Flash, is cheaper and more token-efficient than its predecessor. But the launch is defined as much by what’s missing, the long-delayed flagship Gemini 3.5 Pro, which is reportedly months behind schedule as rivals race ahead.

Key Takeaways

• Google released Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber
• 3.6 Flash is the workhorse, using up to 17% fewer tokens
• 3.5 Flash Cyber is limited to governments and trusted partners
• The flagship Gemini 3.5 Pro is still delayed and not public
• Google’s Pro line hasn’t been updated since February

What Google Released

The drop came Tuesday. Google DeepMind released Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber, a trio aimed at efficiency, latency, and reliability for customers building AI agents at scale.

The lineup splits by job. Gemini 3.6 Flash is the workhorse for coding, knowledge work, and multimodal tasks, 3.5 Flash-Lite is the most cost-effective option in the class, and 3.5 Flash Cyber is a specialized model fine-tuned for finding and fixing security flaws.

The framing is about scale, not frontier bragging rights. Google said the focus is delivering efficiency and reliability to developers deploying agents broadly, positioning these as practical tools rather than a headline-grabbing capability leap.

The Workhorse: Gemini 3.6 Flash

The main event is 3.6 Flash. It promises improved coding, knowledge work, and multimodal performance while reducing token usage by up to 17%, making it cheaper to run than its predecessor, 3.5 Flash.

Efficiency is the selling point. Google says the model takes fewer reasoning steps and tool calls to complete multi-step workflows, which matters for agentic tasks where costs compound across many calls.

It also gained a capability. Like the Lite version, 3.6 Flash now supports computer use as a built-in tool, letting it take on more agentic work, and it ships with added safeguards around chemical, biological, radiological, and nuclear, or CBRN, and cyber-offense misuse.

The Specialist: Gemini 3.5 Flash Cyber

The most restricted model is the cyber one. Gemini 3.5 Flash Cyber is tuned to find and fix cybersecurity vulnerabilities and will be available exclusively to governments and trusted partners through a limited pilot at first.

It’s already at work internally. According to Google DeepMind, the model works alongside an infrastructure agent called CodeMender and is already finding and patching bugs in Google’s own Android, Chrome, and YouTube codebases.

The gated rollout follows a pattern. The limited release mirrors how Anthropic and OpenAI handled their cyber-capable models, Mythos 5 and GPT-5.6 Sol, restricting access because the same skills that defend systems can also be misused to attack them.

The Elephant in the Room: No 3.5 Pro

Here’s what everyone noticed. The launch is notable not just for what shipped but for what it didn’t include, an update to Gemini Pro, Google’s flagship, which was last refreshed back in February.

Google had teased it repeatedly. When it launched 3.5 Flash in May, the company said the Pro version was already being used internally and would roll out the following month, a timeline that has clearly slipped.

Reports point to trouble behind the scenes. Bloomberg reported that Google is facing internal delays launching 3.5 Pro as it struggles to meet its own performance goals, and the company now says only that the model is in partner testing and will ship when ready.

Falling Behind the Frontier

The delay stings because of the competition. Since Google’s last Pro update in February, OpenAI has released GPT-5.5 and begun rolling out GPT-5.6, while Anthropic launched Claude Opus 4.8 and Sonnet 5 and expanded access to its frontier Fable 5, an intense release pace from rival labs.

Even the middle tier is crowded. Chinese labs like Moonshot with Kimi K3 and Zhipu with GLM-5.2 are closing in on frontier performance, and Meta recently shipped a model that outperforms Google’s current lineup at writing code, squeezing Google from multiple directions.

Google is pointing further ahead. Rather than dwelling on the gap, the company has been hyping its next-generation Gemini 4, already in training, a signal it may be trying to leap past the troubled 3.5 Pro cycle rather than salvage it.

Why It Matters

The releases show Google competing where it can. Cheaper, faster, more efficient Flash models are genuinely useful for developers building agents at scale, and token efficiency is exactly the cost-conscious pitch sweeping the industry right now.

But the absence tells the bigger story. For a company that helped invent the transformer, going five months without a flagship refresh while rivals ship frontier models monthly raises real questions about whether Google can keep pace at the top of the market.

The stakes ride on 3.5 Pro, or Gemini 4. Flash models keep developers engaged, but the highest-value reasoning and coding work flows to the strongest frontier models, and until Google ships one, it risks ceding that ground. For now, it’s competing on price and efficiency while the race for raw capability moves on without it.

Digital Trendings is your trusted source for AI news and updates, stay tuned for more.