- Published on
Google announces Gemini 3.6 Flash and cybersecurity AI, teases 3.5 Pro and Gemini 4
- Authors
- Name
- DeepHorus
Google has announced three new AI models, including Gemini 3.6 Flash and a cybersecurity-focused variant, while teasing future releases like Gemini 3.5 Pro and Gemini 4. The announcements come just months after the company unveiled Gemini 3.5 Flash at its I/O conference in May, which has now been deprecated. The new models aim to address user feedback on earlier versions and growing concerns over AI token costs in enterprise deployments.
At Google I/O, Gemini 3.5 Flash was highlighted as a flagship model, but it did not fully meet expectations, particularly in code generation. The company has since shifted focus to efficiency improvements, leading to the release of Gemini 3.6 Flash. This model shows notable benchmark gains: in the DeepSWE test for coding, it jumps to 49 percent from 37 percent for 3.5 Flash. Additionally, the OSWorld test for computer use shows an increase to 83 percent from 78.4 percent, and the model uses about 17 percent fewer tokens.
Google has also introduced Gemini 3.5 Flash Lite and Gemini 3.5 Flash Cyber. The Flash Lite is touted as Google's most efficient modern AI, achieving 350 tokens per second and nearly matching frontier models from a year ago. Its pricing is set at $0.30 per million input tokens and $2.50 per million output tokens. The Flash Cyber model marks Google's first AI specifically designed for cybersecurity applications, though details on its capabilities remain sparse.
Efficiency improvements in Gemini 3.6 Flash translate directly to lower costs for developers. The new model has an API pricing of $1.50 per million input tokens and $7.50 per million output tokens, compared to $1.50 and $9 for 3.5 Flash. This reduction is significant for businesses running large-scale agentic workflows, where token savings can accumulate quickly. Google claims the model completes tasks more accurately and in fewer steps, further reducing operational expenses.
The release of a cybersecurity AI model signals Google's intent to expand into specialized domains. By offering a dedicated model for security tasks, the company aims to attract enterprises that require tailored AI solutions. This move aligns with broader industry trends where AI is increasingly used for threat detection and response.
Industry analysts note that Google is responding to market pressures for cheaper, more capable AI models. The deprecation of Gemini 3.5 Flash just months after its launch highlights the fast pace of iteration in the AI space. However, some developers express cautious optimism, citing the need for consistent performance improvements amid rapid model turnover.
Looking ahead, Google has hinted at the delayed Gemini 3.5 Pro and a forthcoming Gemini 4. These models are expected to push boundaries further, though no release dates have been provided. The company's focus on efficiency and cost reduction suggests that future models will continue to optimize for enterprise use cases, potentially reshaping the competitive landscape against rivals like OpenAI.
