Google Brings Gemini 3.5 Flash-Lite to Search to Power High-Speed Agentic Workflows

Google has introduced a suite of new AI models designed to enhance speed, efficiency, and specific workflow capabilities, with one of the key updates directly impacting how search functions. On July 21, 2026, the company officially announced the launch of three new models: Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber.

Of these three additions, only Gemini 3.5 Flash-Lite is being integrated into Google Search, targeting high-volume, low-latency, and agentic workflows.

Gemini 3.5 Flash-Lite Powers Google Search

Gemini 3.5 Flash-Lite is built specifically to address the low-latency and high-speed requirements of agentic search tools. It is characterized as the fastest and cheapest option within the Gemini model family. Data from the Artificial Analysis Index indicates the model achieves output speeds of 350 tokens per second, representing a speed improvement over earlier Flash-Lite iterations. This high-speed performance is aimed at facilitating multi-step agentic workflows that require fast planning, tool utilization, and sequential reasoning.

Google confirmed that Gemini 3.5 Flash-Lite will roll out within Google Search, though the company did not specify the exact search surfaces that will feature this model. While Google's official announcement links the integration closely to agentic search—which lets users create and manage AI agents to execute multi-step tasks directly inside the search engine—it is possible the model could also be used to power features like AI Overviews and AI Mode. Liz Reid, the head of Google Search, noted that this integration represents a shift in search capabilities.

Aside from its implementation in Search, Gemini 3.5 Flash-Lite is available to developers through the Gemini API in both Google AI Studio and Android Studio.

Performance and Benchmarks

According to internal testing results shared by Google, Gemini 3.5 Flash-Lite outperforms Gemini 3 Flash on key benchmarks:

• SWE-Bench Pro (Coding): Gemini 3.5 Flash-Lite scored 54.2%, compared to 49.6% achieved by Gemini 3 Flash.
• OSWorld-Verified (Computer Task Operation): Gemini 3.5 Flash-Lite scored 74.0%, while Gemini 3 Flash scored 65.1%.

These figures reflect Google's own testing and have not been externally confirmed, but they highlight the performance gains claimed for the new model.

Gemini 3.6 Flash: The Efficiency Workhorse

While Gemini 3.5 Flash-Lite handles search, Google is positioning Gemini 3.6 Flash as its primary workhorse model. Gemini 3.6 Flash is designed to surpass Gemini 3.5 Flash in coding, general knowledge, and multimodal performance, while maintaining high operational efficiency.

According to the Artificial Analysis Index, Gemini 3.6 Flash consumes 17% fewer output tokens compared to Gemini 3.5 Flash. It is capable of completing multi-step workflows with fewer reasoning steps and tool calls. Google has priced Gemini 3.6 Flash at $1.50 per million input tokens and $7.50 per million output tokens, making it more cost-effective than Gemini 3.5 Flash, which costs $9 per million output tokens.

Gemini 3.6 Flash is available via the Gemini Enterprise agent platform, the Gemini Enterprise app, and to developers through the Gemini API in Google AI Studio and Android Studio.

Gemini 3.5 Flash Cyber for Vulnerability Detection

The third model introduced, Gemini 3.5 Flash Cyber, will not be integrated into Google Search or released to the general public. Built on top of Gemini 3.5 Flash, this model is custom-tailored to identify, validate, and patch software vulnerabilities.

Google is restricting access to Gemini 3.5 Flash Cyber through a limited pilot program. It is exclusively available to governments and trusted partners through CodeMender, Google's managed code security service. In testing conducted on complex codebases like Chrome and Safari, Gemini 3.5 Flash Cyber identified more unique vulnerabilities than Gemini 3.5 Flash, Gemini 3.6 Flash, and a rival model developed by Anthropic. Due to the potential risks associated with security-focused AI, Google's security team has implemented strict guardrails to prevent abuse.