Google launches Gemini 3.6 Flash, Flash-Lite and Cyber models

Google has released three new Gemini models aimed at different parts of the AI market: Gemini 3.6 Flash for general work, Gemini 3.5 Flash-Lite for high-volume tasks and Gemini 3.5 Flash Cyber for software security. The company announced the models on July 21. Flash and Flash-Lite are available immediately across Google’s consumer, developer and enterprise channels, while the Cyber model will enter a restricted pilot.

Gemini 3.6 Flash is the new workhorse. Google says it improves coding, knowledge work and multimodal understanding while using fewer tokens than 3.5 Flash. On the Artificial Analysis Index, it reportedly produces 17 percent fewer output tokens, and Google claims larger reductions on some engineering tests. API pricing is $1.50 per million input tokens and $7.50 per million output tokens. The input price is unchanged from 3.5 Flash, but output is cheaper.

Gemini 3.5 Flash-Lite targets workloads where speed and volume matter more than maximum reasoning depth, such as document processing, agentic search and repeated production tasks. Google calls it its fastest and cheapest model in the 3.5 family, citing a speed of 350 output tokens per second. It costs $0.30 per million input tokens and $2.50 per million output tokens. Google’s evaluations show gains over 3.1 Flash-Lite and, on some agent and coding tests, the larger Gemini 3 Flash. These remain provider-selected results and need independent testing in real applications.

The third model has a different risk profile. Gemini 3.5 Flash Cyber is fine-tuned to find, validate and patch software vulnerabilities. Google’s CodeMender system calls several Cyber agents and combines their findings into one report. Because the same skills could be misused offensively, access will initially be limited to governments and trusted partners through CodeMender. Google says the system performed competitively on CyberGym and found previously undisclosed flaws in production code, but it is not yet a generally available security product.

For users, 3.6 Flash and Flash-Lite are rolling out through the Gemini app, Gemini API, AI Studio and enterprise products; Flash-Lite is also coming to Google Search. For makers and businesses, the practical news is lower inference cost and more choice between quality, latency and scale. It also shows model portfolios becoming specialised: one model for daily work, another for enormous task volumes and a guarded one for sensitive capabilities. Gemini 3.5 Pro is still being tested with partners, while Google says Gemini 4 has begun pre-training.