Google ships Gemini 3.6 Flash, Flash-Lite, and Cyber — 3.5 Pro still “when ready”

Google did not ship the frontier model everyone wanted. It shipped the workhorses instead. On July 21, the Gemini team announced Gemini 3.6 Flash, 3.5 Flash-Lite, and a cybersecurity-tuned 3.5 Flash Cyber — while admitting Gemini 3.5 Pro is still in partner testing and will arrive “as soon as it’s ready.” Buried under the Flash fireworks: Gemini 4 pretraining has already started.

What actually shipped

3.6 Flash is the new default workhorse: better coding, knowledge work, and multimodal results, with Artificial Analysis clocking ~17% fewer output tokens than 3.5 Flash (and deeper cuts on some agentic benches). Price: $1.50 / $7.50 per million input/output tokens. 3.5 Flash-Lite is the throughput play — 350 output tokens per second at $0.30 / $2.50 — aimed at high-volume agent loops. Flash Cyber pairs with CodeMender for vulnerability find-and-fix, but only for governments and trusted partners in a limited pilot.

The Pro-shaped hole

3.5 Pro was supposed to land in June after I/O. It did not. Trade coverage has repeatedly tied the slip to coding competitiveness against GPT-5.6-class and Anthropic frontier models. Google’s public line is partner testing only. Until Pro is public, Flash is carrying the brand — and that is a deliberate efficiency story, not a coincidence.

Official post: Introducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber.

Agents got cheaper overnight. The frontier fight still has an empty chair with Google’s name on it.