Introducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber

Details

Google DeepMind launched three Gemini models aimed squarely at production agentic deployments. Gemini 3.6 Flash is the flagship: it delivers better coding and knowledge-work quality than 3.5 Flash while using 17% fewer output tokens on average — and up to 65% fewer on agentic benchmarks like DeepSWE. Pricing drops to $1.50/1M input and $7.50/1M output tokens, lower than 3.5 Flash, which directly reduces the per-task cost of running agents at scale. The model also takes fewer tool calls to complete multi-step workflows. Gemini 3.5 Flash-Lite targets latency-sensitive workloads, hitting 350 output tokens per second on the Artificial Analysis Index — the fastest model in the 3.5 class. Google says it significantly outperforms prior Flash-Lite generations in agentic settings. Gemini 3.5 Flash Cyber is a fine-tuned cybersecurity variant designed to find, validate, and patch software vulnerabilities. It integrates into CodeMender, Google's code security agent, which invokes the model multiple times per scan to cover more code paths. Google says it found 55 unique confirmed issues in the V8 JavaScript engine versus 47 for Gemini 3.5 Flash and 36 for Opus 4.6. It is currently restricted to governments and trusted partners. Buried near the end: Google confirmed it has started its 'most ambitious pre-training run yet' for Gemini 4, signaling the next generation is in motion.