Google Launches Gemini 3.5 Flash Cyber to Close Gap with Anthropic

Google targets cybersecurity with new specialized model

Alphabet released three new Gemini models Tuesday, headlined by Gemini 3.5 Flash Cyber—a specialized system designed to detect and patch software vulnerabilities. The model represents Google's most direct response yet to Anthropic's Mythos, which has established an early lead in automated code defense.

Gemini 3.5 Flash Cyber will initially be available only to government agencies and trusted partners through a limited-access pilot program. Google said the model operates at a lower cost per token than larger general-purpose systems, making it economically viable for high-volume security scanning.

The release comes as Google faces mounting pressure to demonstrate progress across a product pipeline that has experienced delays while competitors gain ground.

Efficiency gains across the model family

Alongside the cybersecurity-focused release, Google launched Gemini 3.6 Flash, which delivers improved performance in coding, multimodal tasks, and knowledge work while using up to 17 percent fewer tokens than its predecessor. The model also costs less per token—a significant reduction for organizations running large-scale workloads.

Gemini 3.5 Flash-Lite rounds out the trio as Google's fastest and least expensive model in the 3.5 family. It's designed for high-volume operations and smaller tasks within larger AI agent systems.

According to Artificial Analysis data, Gemini Flash already undercuts comparable models from Anthropic, OpenAI, and Chinese competitors on cost. Google said Gemini 3.6 Flash—the strongest of the new releases—is cheaper per task than GPT-5.6 Terra Max, Kimi K3, and Qwen 3.7 Max.

Why it matters

Google's emphasis on cost efficiency and specialized models reflects a strategic bet that price and performance can compensate for slower timing in key product categories. As Chinese AI companies like Moonshot AI and Alibaba gain momentum—with Kimi K3 facing capacity constraints due to demand and Alibaba teasing Qwen 3.8 Max—the competitive landscape is intensifying. Building a capable model is only half the challenge; companies must also maintain sufficient computing capacity to serve it at scale. Google's custom chip development and integrated hardware-software approach could provide an advantage, though the company has faced its own capacity limitations.

Hardware integration and future roadmap

The model launches arrive as Google reportedly develops a specialized chip designed to run Gemini up to 10 times more efficiently, part of a broader effort to reduce AI serving costs. A Google Cloud spokesperson told CNBC that the company's teams "constantly research and experiment with new innovations" and that "rigorous exploration is central to our full stack approach."

"By co-designing our hardware and software from the ground up, we ensure our systems are integrated and highly optimized for real-world workloads," the spokesperson added.

Google is also providing greater visibility into its development timeline after questions about delays. Gemini 3.5 Pro is currently being tested with partners ahead of broader availability, while the company has begun its largest-ever pretraining run for Gemini 4.

The releases come on the eve of Alphabet's earnings report, as reported by CNBC.