Google expands Gemini product line, launches cheaper models and Mythos rivals
Alphabet unveiled three new Gemini models on Tuesday, including its clearest response to Anthropic's leadership in cybersecurity. The company is looking to demonstrate its progress as it faces product line delays and increased competition. Gemini 3.5 Flash Cyber is designed to detect and patch software vulnerabilities and will initially be available only to governments and trusted partners through a limited-access pilot program. Said that compared with large models, the price per word of this special model is lower. This helps Google close the gap in cybersecurity with Anthropic, which has established an early lead in automated code defense. Google also launched Gemini 3.6 Flash, which has improved encoding, multi-modal and knowledge work performance, while using up to 17% fewer tokens than the previous generation model and a lower unit cost per token - which means a significant cost reduction for running high-throughput workloads. Gemini 3.5 Flash-Lite, meanwhile, is the fastest and lowest-cost model in Google's 3.5 series, built for high-throughput workloads and smaller tasks in large AI agent systems. The broader product line reflects Google's bet that price and efficiency will help offset its slower rollouts in several key product categories. Artificial Analysis data shows that Gemini Flash is already lower in cost than comparable models from Anthropic, OpenAI and Chinese competitors. According to Google, the more capable of the two new models, the Gemini 3.6 Flash, costs less per task than the GPT-5.6 Terra Max, Kimi K3, and Qwen 3.7 Max, while the 3.5 Flash-Lite costs a fraction of that. This demand highlights another side of the AI race: Building competitive models is only part of the challenge; companies also need enough computing power to deliver services at scale. Google has a potential advantage through its ability to customize chips, cloud infrastructure and co-engineer software and hardware, although the company has faced capacity constraints of its own. Tuesday's model launch comes as Google is reportedly developing a specialized chip designed to make Gemini run up to 10 times more efficiently, as part of its broader effort to lower the cost of artificial intelligence services. A Google Cloud spokesperson said in a statement that its team "continues to research and try new innovations to deliver maximum performance and efficiency for our users and customers" and that "while not every project will make it to production, this rigorous exploration is at the core of our full-stack approach." "By co-engineering our hardware and software from the ground up, we ensure our systems are integrated and highly optimized for real-world workloads," the statement continued. Google is also making its product roadmap more transparent after facing questions about delays. Gemini 3.5 Pro is currently being tested with partners in preparation for wider release, while the company has launched its largest-ever Gemini 4 pre-training run.