Chrome OS Update: 3.6 Flash & 3.5 Flash Variants Available for Enthusiasts & Power Users

Key Points

• Gemini 3.6 Flash reduces output tokens by 17% compared to 3.5 Flash and lowers costs for AI workflows, making it more efficient for developers.
• 3.5 Flash-Lite is Google’s fastest and cheapest option, delivering high-speed performance for cost-sensitive applications.
• CodeMender’s Cyber model introduces a security-focused AI agent to enhance cybersecurity workflows, with plans for expanded Gemini Pro and Gemini 4 models.

Google has unveiled new AI models designed to improve efficiency and affordability for developers building agentic workflows. The Gemini 3.6 Flash model, an upgraded version of its predecessor, focuses on reducing costs and computational demands. It uses 17% fewer output tokens than Gemini 3.5 Flash, according to benchmarks, while also cutting prices at $1.50 per million input tokens and $7.50 per million output tokens. This model prioritizes coding, knowledge-based tasks, and multimodal capabilities, enabling faster and cheaper AI-driven applications.

For speed-focused use cases, Gemini 3.5 Flash-Lite emerges as the fastest option in its class, delivering up to 350 output tokens per second. It excels in balancing cost and performance, particularly for scalable AI agents. Meanwhile, Google introduced CodeMender’s Cyber model as part of its cybersecurity toolkit. This specialized model pairs with an agent infrastructure to streamline security tasks like code analysis, offering competitive performance while addressing privacy and compliance needs.

Future developments include Gemini 3.5 Pro, currently in testing with partners, and Gemini 4, which Google is pre-training with what it calls its “most ambitious run yet.” These advancements aim to meet growing demands for reliable, low-latency AI in complex workflows.

Developers and businesses leveraging these models will benefit from lower operational costs and faster deployment timelines. For example, 3.6 Flash reduces the number of “reasoning steps” required for multi-step tasks, a critical win for agentic systems that often depend on iterative processing. Early benchmarks, such as Datacurve’s DeepSWE evaluation, show up to 65% token savings, which could significantly cut cloud computing expenses.

The Flash-Lite variant addresses cost-sensitive projects, from chatbots to real-time data processing, where speed and budget constraints matter most. Its performance gains over earlier Flash-Lite models underscore Google’s focus on iterative improvements tailored to specific use cases.

The CodeMender Cyber model targets a niche but critical space: cybersecurity. By embedding specialized expertise into AI workflows, Google aims to automate tasks like vulnerability detection while maintaining high accuracy. The integration of an “agent infrastructure” suggests a layered approach, where the model works alongside rule-based systems or other tools—a strategy common in enterprise-grade AI.

Looking ahead, Gemini 3.5 Pro promises to bridge the gap between efficiency and advanced reasoning capabilities. While details remain under wraps, its broader availability signals Google’s readiness to support enterprise and research applications. The upcoming Gemini 4 represents a leap into more powerful pre-training, hinting at future breakthroughs in natural language processing and multimodal interaction.

These updates reflect Google’s push to make AI tools accessible to a wider range of developers—from startups to large enterprises—by refining both cost structures and technical performance. For organizations building production agents, the new models offer a clearer path to scalability without sacrificing quality.

As AI agents become integral to business workflows, Google’s focus on “token efficiency” and “latency reduction” positions its Gemini lineup as a pragmatic choice for teams prioritizing reliability and cost-effectiveness. The next steps, particularly Gemini 4, could redefine how AI is deployed across industries.

Read the rest of the article

You can also check out our list of the Best Instagram Extensions, Best Pinterest Exensions & the Best AI Extensions.