Google ships cheaper Gemini 3.6 Flash and 3.5 Flash-Lite, plus a government-only cyber model

Google released three new Gemini AI models on July 21, 2026, led by Gemini 3.6 Flash, a cheaper, faster version priced at $1.50 per million input tokens, the company said.

The $7.50 per million output-token price is down from $9 on the prior 3.5 Flash, and Google says the new model uses 17 percent fewer output tokens on the same tasks — a compounding discount for developers who run high volumes. On benchmarks Google cites, 3.6 Flash scores 83.0 percent on OSWorld-Verified, up from 78.4 percent, and improves as much as 65 percent on the DeepSWE coding test.

The lineup also includes Gemini 3.5 Flash-Lite, a lower-cost model at $0.30 per million input and $2.50 per million output tokens that runs at 350 output tokens per second, and Gemini 3.5 Flash Cyber, a cybersecurity variant offered only to governments and trusted partners through Google’s CodeMender program.

Both 3.6 Flash and 3.5 Flash-Lite went live immediately across the Gemini API, Google AI Studio, Android Studio, Gemini Enterprise and the Gemini app. The same day, GitHub Copilot began rolling out 3.6 Flash to its Pro, Pro+, Max, Business and Enterprise tiers.

The benchmark gains are Google’s own numbers and have not been independently verified; real-world coding and agent performance often diverges from eval scores. The release also lands with Google’s flagship Gemini Pro model still delayed, leaving the company competing on price and speed at the low end while the high end waits.

Whether the government-only Cyber variant stays locked down, or Google widens access as rivals ship security models, is the signal to watch.