Google has expanded its artificial intelligence model portfolio with Gemini 3.6 Flash, Gemini 3.5 Flash-Lite and Gemini 3.5 Flash Cyber, targeting cost-efficient agentic workloads, high-volume processing and automated software security.
The releases give developers and enterprises more specialized options for balancing model intelligence, speed and operating costs. However, Google’s anticipated Gemini 3.5 Pro model remains under testing with partners and is expected to arrive when it is ready.
TL;DR
- Gemini 3.6 Flash improves coding, knowledge work and token efficiency while reducing the cost of agentic tasks.
- Gemini 3.5 Flash-Lite delivers 350 output tokens per second for high-volume, low-latency workloads.
- Gemini 3.5 Flash Cyber finds, validates and patches vulnerabilities through Google’s CodeMender security agent.
Gemini 3.6 Flash: Better Quality At A Lower Cost
Gemini 3.6 Flash builds on Gemini 3.5 Flash, with Google positioning it as a more efficient model for coding, knowledge work, multimodal analysis and complex multi-step workflows.
According to Google, Gemini 3.6 Flash consumes 17% fewer output tokens than Gemini 3.5 Flash on the Artificial Analysis Index. It also needs fewer reasoning steps and tool calls to complete multi-step workflows.
The model is priced at $1.50 per one million input tokens and $7.50 per one million output tokens. It scored 1,421 on the GDPval-AA v2 knowledge-work benchmark, compared with 1,349 for Gemini 3.5 Flash.
Google said the model improves document parsing, chart and data analysis, report drafting, code migration and visual development. It also ships with enhanced safeguards covering chemical, biological, radiological and nuclear risks, as well as cyber-offense misuse.
Gemini 3.5 Flash-Lite: Built For High-Volume Workloads
Gemini 3.5 Flash-Lite is Google’s fastest and most cost-effective model in the Gemini 3.5 series. As measured by Artificial Analysis, it generates 350 output tokens per second.
It is priced at $0.30 per one million input tokens and $2.50 per one million output tokens, making it suitable for agentic search, document processing, receipt translation and large-scale data extraction.
The model scored 54% on Terminal-Bench 2.1, compared with 31% for Gemini 3.1 Flash-Lite. It also reached 72.2% on Google DeepMind’s long-context MRCR v2 evaluation and 1,140 on GDPval-AA v2.
Google added configurable thinking levels, allowing developers to prioritize lower latency and costs or allocate more reasoning capacity to multi-step subagent tasks. Computer use is also available as a built-in tool.
Gemini 3.5 Flash Cyber Targets Software Vulnerabilities
Gemini 3.5 Flash Cyber is a specialized version of Gemini 3.5 Flash, fine-tuned to find, validate and patch software vulnerabilities. It powers CodeMender, where multiple Flash Cyber agents inspect different code paths and combine their findings into one report.
In tests involving the V8 JavaScript engine, Flash Cyber found 55 unique confirmed issues. This compared with 47 discovered by the standard Gemini 3.5 Flash model and 36 found by Claude Opus 4.6, including 10 vulnerabilities missed by both models.
Google said its Cloud Vulnerability Research team used the model to uncover remote-code-execution vulnerabilities and a memory-corruption issue in a production service within two hours.
“By building on top of Flash, 3.5 Flash Cyber offers a cost-efficient and highly capable alternative to large, costly cybersecurity models,” Google DeepMind researchers Raluca Ada Popa and Four Flynn said.
Given the technology’s dual-use risks, Flash Cyber will initially be offered only to governments and trusted partners through a limited-access CodeMender pilot.
Gemini 3.6 Flash and Flash-Lite are available through the Gemini API, Google AI Studio, Android Studio, Gemini Enterprise and the Gemini app. Flash-Lite is also rolling out in Google Search, while Google says Gemini 3.5 Pro is coming soon.
