What's this about?
DeepMind refreshed its "Flash" line — the small, fast, cheap Gemini models — with a new 3.6 Flash, an even lighter 3.5 Flash-Lite, and a security-specialized "3.5 Flash Cyber."
What exactly is a "Flash" model?
Labs ship a big flagship model and smaller "fast/cheap" siblings for high-volume, latency-sensitive work (autocomplete, classification, bulk processing). "Flash" is Gemini's name for that tier. "Flash Cyber" being a distinct variant suggests a model tuned for cybersecurity tasks specifically.
Why it matters
Most real production traffic runs on the cheap tier, not the flagship — improvements here move more actual workloads than a headline model does.