Google DeepMind releases three new Gemini models: 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber

Google DeepMind announced three new models in the Gemini series—Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber—each designed to address specific needs in efficiency, latency, and reliability for building AI agents at scale.

Gemini 3.6 Flash is positioned as a workhorse model, delivering improvements in coding, knowledge work, and multimodal performance. According to the Artificial Analysis Index, it reduces output token usage by 17% compared to 3.5 Flash, and in some benchmarks like DeepSWE by Datacurve, it achieves up to 65% fewer output tokens at a lower cost per output token. The model is priced at $1.50 per 1M input tokens and $7.50 per 1M output tokens, making it more cost-effective for agentic tasks.

Gemini 3.6 Flash also demonstrates performance gains over 3.5 Flash across several benchmarks, including higher precision in code edits (49% vs. 37% in DeepSWE), improved ML research capabilities (63.9% vs. 49.7% in MLE Bench), and better computer use performance (83.0% vs. 78.4% in OSWorld-Verified). The model's computer use capabilities are now available as a built-in client-side tool via the Gemini API and Gemini Enterprise.

Gemini 3.5 Flash-Lite is designed for low-latency and high-throughput tasks, such as agentic search and document processing. It delivers 350 output tokens per second according to the Artificial Analysis Index and significantly outperforms prior Flash-Lite generations in agentic workflows.

Gemini 3.5 Flash Cyber is a specialized model for cybersecurity applications, introduced alongside the CodeMender code security agent. This combination is intended to deliver competitive performance at the frontier for cybersecurity use cases.

The updates also include enhanced Frontier Safety safeguards for Gemini 3.6 Flash, particularly in domains such as Chemical, Biological, Radiological, and Nuclear (CBRN) and cyber offense misuses, making the model more resistant to jailbreaks while minimizing refusals for beneficial uses.