The Zhitong Finance App learned that on the eve of the financial report, Google, a subsidiary of Alphabet (GOOGL.US, GOOG.US), officially launched three new Gemini models on Tuesday, covering various fields such as cybersecurity, programming, and cost-effective reasoning. The aim is to show the latest developments in the AI product line and cope with the increasingly fierce competition of Anthropic and other AI companies.
The products released include Gemini 3.5 Flash Cyber, Gemini 3.6 Flash, and Gemini 3.5 Flash-Lite.
Among them, Gemini 3.5 Flash Cyber was developed specifically for network security scenarios and can be used to identify and fix software vulnerabilities. Initially, it will only be open to government agencies and trusted partners through limited access testing. Google said that while the model maintains professional capabilities, the cost of using a single token is lower than the larger scale model.
The industry believes that this product is expected to help close the gap between Google and Anthropic in the field of cybersecurity AI. Previously, Anthropic was the first to establish a leading edge in the automated code security market with the Mythos model.
Meanwhile, Google also launched Gemini 3.6 Flash. The company said that the new model has further improved the performance of programming, multi-modal processing and knowledge work tasks. At the same time, token consumption has been reduced by up to 17% compared to the previous generation, and the cost per token has been further reduced, which helps reduce the cost of deploying large-scale AI applications.
Another Gemini 3.5 Flash-Lite is positioned as the fastest and lowest-cost model in the Gemini 3.5 series. It is mainly aimed at highly concurrent tasks and lightweight workloads in AI agent systems.
Analysts believe that the current product matrix shows that Google is trying to make up for its lag in the launch time of some AI products through lower costs and higher efficiency.
According to Artificial Analysis data, the Gemini Flash series model is currently lower than similar products from Anthropic, OpenAI, and some other AI companies in terms of price. According to Google, the new Gemini 3.6 Flash, which has the strongest performance, is also lower than OpenAI's GPT-5.6 Terra Max, Kimi K3 from Dark Side of the Moon, and Alibaba (BABA.US)'s Qwen 3.7 Max in terms of unit task costs.
This new product launch comes at a time when Alphabet is about to announce quarterly results. At the same time, competition among AI companies continues to heat up.
Recently, Kimi K3, which was launched on Dark Side of the Moon, is in high demand, and the company temporarily restricted new subscriptions and API access due to insufficient computing power. At the same time, Alibaba also predicted the upcoming launch of Qwen 3.8 Max, and indicated that its overall performance is second only to Anthropic's latest flagship model, Fable 5.
The industry believes that this also highlights another key factor in the current AI competition. Building a leading model is only the first step; enterprises must also have sufficient computing power infrastructure to support large-scale commercial deployment of models.
In this field, Google has certain advantages with self-developed AI chips (TPU), cloud infrastructure, and software and hardware co-design capabilities, but the company also faced tight computing resources before.
It is worth mentioning that, according to previous media reports, Google is currently developing a new AI chip specifically designed to run the Gemini model. The goal is to increase operating efficiency by up to 10 times and further reduce the cost of AI services.
A Google Cloud spokesperson said in a statement that the company team continues to explore and test new technologies to provide users with higher performance and more efficient AI services. Although not all R&D projects end up in production, this continuous innovation is an important part of Google's full-stack AI strategy.
The spokesperson said that by collaborating to design hardware and software from the bottom up, Google can build an AI system that is highly integrated and optimized for actual workloads.
Furthermore, in response to market concerns about product delays, Google also disclosed more product roadmap information for the first time. The company said Gemini 3.5 Pro has now begun open testing with partners to prepare for the full launch. At the same time, Gemini 4 has also begun the largest pre-training effort to date.