Google expands Gemini Flash lineup with 3.6 Flash, 3.5 Flash-Lite and Flash Cyber model

Google has expanded its Gemini family with a new lineup of Flash models aimed at scaling agentic AI workflows across coding, enterprise and cybersecurity use cases. Leading the announcement is Gemini 3.6 Flash, the company's latest workhorse model, which delivers improved coding, knowledge work and multimodal performance. According to the Artificial Analysis Index, it reduces output token usage by 17% compared with Gemini 3.5 Flash, while benchmarks such as DeepSWE by Datacurve show token savings of up to 65%, all at a lower cost per output token.

Google also introduced Gemini 3.5 Flash-Lite, its fastest and most cost-effective model in the 3.5 series. The model can generate up to 350 output tokens per second, according to the Artificial Analysis Index, and significantly outperforms previous Flash-Lite generations in agentic workflows, making it well suited for high-throughput AI tasks.

More from Tech

Rounding out the announcements is Gemini 3.5 Flash Cyber, a specialised model built for cybersecurity applications. Designed to work alongside CodeMender, Google's code security agent, it combines a cyber-focused AI model with an agentic security framework to help identify and address software security vulnerabilities more efficiently.

How is Gemini 3.6 Flash vs 3.5 Flash?

Gemini 3.6 Flash builds directly on developer and customer feedback from Gemini 3.5 Flash. While the new model delivers improvements in coding and knowledge work, Google says its biggest advancement lies in token efficiency. According to the Artificial Analysis Index, Gemini 3.6 Flash consumes 17% fewer output tokens than Gemini 3.5 Flash. It also requires fewer reasoning steps and tool calls to complete multi-step workflows.

The new model is priced at $1.50 per million input tokens and $7.50 per million output tokens, reducing the overall cost of running agentic AI tasks.

Google also said Gemini 3.6 Flash improves coding accuracy by reducing unwanted code edits and execution loops, scoring 49% on DeepSWE, up from 37% for Gemini 3.5 Flash. It boosts machine learning research performance as well, achieving 63.9% on MLE Bench compared with 49.7% previously. The model further enhances computer-use capabilities, scoring 83.0% on OSWorld-Verified versus 78.4% for its predecessor, while also outperforming Gemini 3.5 Flash in knowledge work with a GDPval-AA v2 score of 1,421, up from 1,349. Google added that the model also offers stronger multimodal capabilities for tasks such as document parsing, chart and data analysis, and report drafting.

Safety features

Gemini 3.6 Flash ships with enhanced frontier safety safeguards across chemical, biological, radiological, nuclear (CBRN) and cyber-offence misuse scenarios. Google said these safeguards make the model more resistant to jailbreak attempts while also reducing unnecessary refusals for beneficial and legitimate use cases.

Gemini 3.5 Flash-Lite focuses on speed

Google said Gemini 3.5 Flash-Lite is the fastest model in the 3.5 series, capable of generating 350 output tokens per second, while being priced at $0.30 per million input tokens and $2.50 per million output tokens. Designed for high-volume, low-latency workloads, the model supports configurable reasoning levels for agentic tasks and includes built-in computer-use capabilities.

According to Google, the model delivers significant performance gains over Gemini 3.1 Flash-Lite, scoring 54% on Terminal-Bench 2.1 (up from 31%), 72.2% on GDM-MRCR v2 (up from 60.1%), and 1,140 on GDPval-AA v2 (up from 642). It also outperforms Gemini 3 Flash on select benchmarks, including SWE-Bench Pro (54.2% vs. 49.6%) and OSWorld-Verified (74.0% vs. 65.1%), making it a faster and more capable option for coding and agentic AI workloads.

CodeMender gets Gemini 3.5 Flash Cyber

Google has also introduced Gemini 3.5 Flash Cyber, a specialised version of Gemini 3.5 Flash fine-tuned to detect, validate and patch cybersecurity vulnerabilities more efficiently while offering a lower cost per token than larger models.

Integrated into CodeMender, the model uses multiple AI agents working together to generate a consolidated security report and delivers competitive performance on the CyberGym benchmark. Owing to the dual-use nature of the technology, Google said Gemini 3.5 Flash Cyber will initially be available only to governments and trusted partners through a limited-access CodeMender pilot programme.

Google said Gemini 3.6 Flash and Gemini 3.5 Flash-Lite are available starting today.