Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber

Google rolled out a fresh trio of AI models — Gemini 3.6 Flash, 3.5 Flash-Lite, and a security-focused Cyber version — promising they’re faster, cheaper, and better for the kind of behind-the-scenes tasks companies use to build AI helpers. On paper, 3.6 Flash is the big star: Google says it uses fewer words to get the job done, costs less, and performs better on coding, document reading, and other work tasks. Flash-Lite is being pitched as the speed demon of the family, while the Cyber model is aimed at security teams.

But the real fireworks were in the comments, where the launch got a collective eyebrow raise instead of a standing ovation. One of the loudest reactions was basically: cool story, but why would I use this? Critics said the benchmark scores didn’t feel exciting, especially because Google mostly compared the new models to its own older models instead of rivals. That drew some blunt side-eye, with one commenter accusing the company of being deeply inward-looking and acting like outside competition barely exists.

Then came the math drama. A commenter pounced on Google’s numbers, asking how the model could be “up to 65%” better in one place but only show “49% vs. 37%” elsewhere. Translation for non-nerds: the crowd smelled possible marketing spin and immediately started fact-checking the fine print. Another hot take went straight for the jugular, claiming the new model is less smart, pricier, and closed off compared to a Chinese competitor. In other words, Google wanted applause for a shiny upgrade — and the internet responded with receipts, sarcasm, and a group chat full of skepticism.