The Gemini Flash Buyer's Sheet — Gemini 3.6 Flash, 3.5 Flash-Lite & Flash Cyber: Which to Use, and the Cost-Per-Task Math the Sticker Hides

The companion to the full video. On July 21, 2026 Google shipped three new models — Gemini 3.6 Flash (better than its predecessor on every published benchmark AND cheaper: output cut from $9.00 to $7.50 per million tokens), Gemini 3.5 Flash-Lite ($0.30 in / $2.50 out, ~350 output tokens per second, and it beats last generation's FULL-SIZE Gemini 3 Flash on SWE-Bench Pro and OSWorld), and the locked Gemini 3.5 Flash Cyber (governments and trusted partners only). THIS SHEET: the 3-rule picker card that ends model-choice debates, the full official price table including the half-price Batch & Flex lane most people never use ($0.75/$3.75 for 3.6 Flash, $0.15/$1.25 for Flash-Lite), the cost-per-finished-task math — sticker says −17%, but with Google's own token-efficiency numbers a typical task lands near −31% and agentic coding near −70% (derivation shown, honestly labeled as derived from vendor benchmarks), the thinking-dial trick on Flash-Lite (minimal / low / higher — thinking IS output tokens), the free-tier warning (Google can train on free-tier content), and a 4-step do-this-today checklist. Every price verified against Google's official pricing page on July 21, 2026. Independent field manual from Hyperautomation Labs.

By submitting, you'll be subscribed to the Hyperautomation AI Report newsletter — one practical AI email per week. Unsubscribe anytime.

Free. No spam. Unsubscribe anytime.