Google Ships Gemini 3.6 Flash, 3.5 Flash-Lite, and Flash Cyber — and Starts Gemini 4 Pre-training

Google launched three Gemini Flash models on July 21, cutting output pricing on its workhorse tier and adding a security-specialist model to the API.

Google shipped three new Gemini Flash variants, all live immediately in the Gemini API and Google AI Studio:

• Gemini 3.6 Flash — improved coding, knowledge work, and multimodal performance. API pricing: $1.50/1M input, $7.50/1M output (down from $9/1M output). Knowledge cutoff extended to March 2026. Available in Google AI Studio, Android Studio, and the Gemini API.
• Gemini 3.5 Flash-Lite — the fastest in the 3.5 family; optimized for throughput and low latency at high volume.
• Gemini 3.5 Flash Cyber — a security-specialist model tuned for finding and fixing vulnerabilities; available in the Gemini Enterprise Agent Platform.

Google also confirmed the start of pre-training for Gemini 4, described as "our most ambitious pre-training run yet," while Gemini 3.5 Pro remains in limited partner testing.

Why it matters: The 3.6 Flash output price cut tightens competition at the budget model tier where most high-volume agent pipelines run. Flash Cyber is the first security-domain-specialized model from a major lab to ship directly in a public API — worth evaluating if you are building automated vuln-scanning or code-review agents.

Sources: Gemini 3.6 Flash announcement — Google Blog · Gemini API changelog — Google AI