Skip to content

AI

Google ships three Gemini models, still no 3.5 Pro in sight

Gemini 3.6 Flash, 3.5 Flash-Lite and a locked-down Cyber variant landed on Tuesday. The flagship Pro everyone keeps asking about stayed home.

Cover photo: Gregory Varnum / CC BY-SA 4.0

TL;DR

  1. Gemini 3.6 Flash cuts token usage by up to 17%, per Google.
  2. Flash-Lite is pitched as the cheapest in its class; Flash Cyber is for governments and partners only.
  3. The long-awaited Gemini Pro update did not show up. Pro was last updated in February.

What happened

On Tuesday, Google DeepMind released Gemini 3.6 Flash, Gemini 3.5 Flash-Lite and Gemini 3.5 Flash Cyber. Three names, three jobs, zero flagship.

Gemini 3.6 Flash is the "workhorse model". Google promises better coding, knowledge work and multimodal performance. It also says the model uses up to 17% fewer tokens than 3.5 Flash, which makes it cheaper to run.

Gemini 3.5 Flash-Lite is billed as the most cost-effective model in its class. That is the budget pick for anyone burning through requests.

Then there is Gemini 3.5 Flash Cyber. It is fine-tuned for finding and fixing cybersecurity vulnerabilities. You cannot just sign up: Google is limiting it to governments and trusted partners in a pilot program.

Google says the aim across all three is efficiency, latency and reliability for customers building AI agents at scale.

Why it matters

The interesting part is what is missing. The long-anticipated update to Gemini Pro, the flagship, did not ship. Pro was last updated in February.

Meanwhile, OpenAI has put out GPT-5.5 and started rolling out GPT-5.6 since then. So Google answered with cheaper, faster workhorses rather than a headline-grabbing big brain.

That is a legitimate strategy. Agents make a lot of calls, and token costs pile up fast. A model that does the same job with fewer tokens is a boring improvement that shows up directly on the invoice.

The Cyber model is the quiet curveball. A security-tuned model that only vetted organizations can touch tells you Google sees both the upside and the risk of handing vulnerability-hunting tools to everyone.

Cheaper workhorses today, flagship later. Google picked the invoice over the fireworks.

What it means for Europe

The general Flash and Flash-Lite models are open to developers in the EU like everyone else. The Cyber variant is not: it is restricted to governments and trusted partners in the pilot. European security teams will need to be on that partner list or keep waiting.

What to do with it

  • Re-run your costs.If you run agents on 3.5 Flash, test 3.6 Flash on your own workload and compare the token bill, since the 17% is a ceiling and a vendor claim.
  • Try Flash-Lite for the boring stuff.Classification, routing and summaries are the obvious candidates for the cheapest tier.
  • Hold your Pro plans.If you were waiting for a new Gemini Pro, nothing changed today, so do not rebuild around a rumor.
  • Security teams, ask early.If you work in government or with a trusted partner, find out how to join the Flash Cyber pilot.

Ciao, sapiens. The workhorse got cheaper. Your coffee budget, sadly, did not.