Google Launches Gemini 3.6 Flash and Flash Cyber, Begins Training Gemini 4

Google drops three new Gemini models: 3.6 Flash with 17% better token efficiency, 3.5 Flash-Lite at 350 tokens/sec, and Flash Cyber for security scanning. Gemini 4 pre-training underway.

Google has launched a trio of new Gemini models alongside news that pre-training for Gemini 4 has begun.

Gemini 3.6 Flash is the new workhorse, delivering better coding, knowledge work, and multimodal performance while consuming 17% fewer output tokens than its predecessor. On the DeepSWE benchmark, token reductions reached 65% in some cases — all at a lower per-token price of $1.50 per million input tokens and $7.50 per million output tokens.

Gemini 3.5 Flash-Lite pushes speed to 350 output tokens per second, making it Google's fastest and most cost-effective 3.5-class model, with significant gains in agentic workflows compared to prior Flash-Lite generations.

The standout release is Gemini 3.5 Flash Cyber, a security-focused model paired with Google's CodeMender agent. It found 55 unique confirmed bugs in Chrome's V8 JavaScript engine — 10 of which no other model discovered — outperforming Gemini 3.5 Flash (47 bugs) and Anthropic's Claude Opus 4.6 (36). The model is designed to compete with Anthropic's larger Mythos system, offering a cost-efficient alternative for vulnerability scanning that can be invoked multiple times at high speed to explore deeper code paths.

Google also confirmed that Gemini 3.5 Pro is currently in partner testing, while the team has started what it calls "our most ambitious pre-training run yet" for Gemini 4.

Sources: Google Blog, The Verge