Google has started rolling out its newest AI model, Gemini 3.5 Flash-Lite, inside Google Search. The company officially announced the launch along with two other models on July 21, 2026. The Flash-Lite release is built specifically for speed and the kind of high-volume, low-latency work that agentic search tools depend on.
The rollout is part of a broader announcement that included three new models: Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber. Among these three models, only 3.5 Flash-Lite will be rolled out to Google Search. The remaining models have different purposes, and Google is clear on who gets access to them.
What Google Announced
According to Google’s presentation, 3.5 Flash-Lite is the fastest and cheapest model of the Gemini family. Based on figures provided by the Artificial Analysis Index, the model provides 350 tokens/second outputs, which is an improvement over previous Flash-Lite models, particularly in agentic workflows requiring fast planning, tool use, and response across several steps.
Google also claimed that 3.5 Flash-Lite significantly outperforms 3 Flash on two established benchmarks. On SWE-Bench Pro, a coding benchmark, the model scored 54.2% against 3 Flash’s 49.6%. n OSWorld-Verified, which measures a model’s ability to operate a computer to complete specific tasks, 3.5 Flash-Lite scored 74.0% against 3 Flash’s 65.1%. It is important to note that the figures come directly from Google’s own announcement and should be viewed as internal testing results rather than an external confirmation. But they still give a sense of the claims the company is making.
Gemini 3.6 Flash, the second model Google introduced, is being positioned as its workhorse model. The company noted that it is superior to 3.5 Flash in coding, knowledge, and multimodal performance while maintaining efficiency. Compared to 3.5 Flash, 3.6 Flash consumes 17% fewer output tokens, per the Artificial Analysis Index. 3.6 Flash requires fewer reasoning steps and tool calls to complete multi-step workflows. As for pricing, the model costs $1.50 per million input tokens and $7.50 per million output tokens, compared to the $9 per million output tokens that 3.5 Flash costs.
The third model, Gemini 3.5 Flash Cyber, will not be rolled out to Google Search or released to the public. Being built upon 3.5 Flash, it is designed for finding, validating, and patching software vulnerabilities. Google is releasing it through a limited-access pilot program available exclusively to governments and trusted partners through CodeMender, its managed code security service. During tests on complex codebases, including Chrome and Safari, 3.5 Flash Cyber detected more unique vulnerabilities than 3.5 Flash, 3.6 Flash, and a competing model from Anthropic. Because of the risks of misusing such technology, Google’s security team added guardrails to the model to reduce potential abuse.
Where 3.5 Flash-Lite Is Showing Up
In practice, the two models actually reaching users right now are Gemini 3.6 Flash and 3.5 Flash-Lite. The former is now available through Gemini Enterprise agent platform and Gemini Enterprise app, while both can be reached by developers through Gemini API in Google AI Studio and Android Studio. Flash-Lite is the version of the model that will be integrated into Google Search.
Google itself said the new model will also roll out in Google Search, though it didn’t specify which particular surfaces within Search will use this version of Gemini. Search Engine Land noted that Google’s blog post about the release ties it to agentic search, but it doesn’t rule out the possibility that this model could also power AI Overviews and AI Mode.
The fact that this model will be integrated into agentic search surfaces becomes particularly interesting as agentic search allows for creating and managing AI agents right inside Google Search in order to complete multi-step tasks. According to Liz Reid, the head of Google Search, this marks a shift toward letting people create, customize, and manage multiple AI agents for various tasks directly within Search. Since Flash-Lite is designed to process information quickly and affordably, deploying it in these surfaces makes sense.
Why This Matters For SEO And Search Marketers
Even though Google hasn’t spelled out every detail, this kind of change to underlying infrastructure is worth watching for anyone who follows how Google Search works. A faster and cheaper model powering AI Overviews and AI Mode would mean faster performance of these features, as well as greater capacity to implement them elsewhere, as well as perform multi-source agentic tasks.
At the same time, this shows that Google keeps iterating actively on its Flash-series models rather than waiting for a single full release. Gemini 3.5 Flash itself only came out in May, and three more models followed within about two months, which suggests Google feels pressure to keep improving speed and cost efficiency across its AI models, particularly the ones powering consumer-facing services.
There is also the larger-scale release to look forward to. At I/O event held in May, Google announced Gemini 3.5 Pro along with its expected release for June. That release never happened on schedule, but Google addressed it in its latest blog post, saying Gemini 3.5 Pro is currently undergoing partner testing and will be released once testing is complete. Besides that, Google said it started the pre-training of Gemini 4. That leaves several Gemini release tracks moving at once, with one version delayed while another ships.
What To Watch Next
Right now, Google hasn’t provided a detailed breakdown of what portion of AI Overviews or AI Mode traffic will run on 3.5 Flash-Lite versus other models, nor a timeline for when this rollout will be complete. It is not clear whether this process is going to be gradual or not.
For marketers and SEO professionals, this is an example of how Google’s underlying infrastructure continues to change even without obvious changes in user experience. Faster models mean more room for quicker iteration and evolution across AI Overviews, AI Mode, and agentic search.
Frequently Asked Questions
Q1: What is Gemini 3.5 Flash-Lite?
It’s Google’s fastest and most cost-effective model in the Gemini 3.5 series, built for low-latency tasks and high-throughput agentic workflows. Google says it delivers 350 output tokens per second and outperforms earlier Flash-Lite versions on agentic benchmarks.
Q2: Is Gemini 3.5 Flash-Lite available in Google Search now?
Yes. Google confirmed on July 21, 2026, that 3.5 Flash-Lite is rolling out in Google Search, alongside its existing rollout in the Gemini app and developer tools.
Q3: Which parts of Google Search will use 3.5 Flash-Lite?
Google specifically tied the rollout to agentic search. The company did not confirm whether AI Overviews or AI Mode will also run on this model, though it hasn’t ruled that out either.
Q4: What other models did Google announce alongside 3.5 Flash-Lite?
Google also announced Gemini 3.6 Flash, a more capable workhorse model for coding and knowledge work, and Gemini 3.5 Flash Cyber, a specialized model for finding and fixing software vulnerabilities.
Q5: Is Gemini 3.5 Flash Cyber available to the public?
No. Google is limiting access to governments and trusted partners through CodeMender, its managed code security tool, as part of a controlled pilot program.
Q6: How does Gemini 3.6 Flash compare to Gemini 3.5 Flash?
Google says 3.6 Flash reduces output token usage by 17% according to the Artificial Analysis Index, takes fewer reasoning steps to complete multi-step tasks, and costs less per output token than 3.5 Flash.
Q7: What is agentic search?
Agentic search refers to Google Search features that let users create and manage AI agents to handle multi-step tasks directly within Search. Google first detailed this at I/O in May 2026.
Q8: What happened to Gemini 3.5 Pro?
Gemini 3.5 Pro was announced at I/O in May with a planned June release but did not ship on that timeline. Google now says it’s in partner testing and will launch broadly once ready.
Q9: Why does this update matter for SEO professionals?
Faster, cheaper models powering Search’s AI features could affect how quickly AI Overviews, AI Mode, and agentic tools evolve, which makes it worth tracking even without a full breakdown of where each model is deployed.
Q10: How can developers access these new models?
Gemini 3.6 Flash and 3.5 Flash-Lite are available now through the Gemini API via Google AI Studio and Android Studio, with 3.6 Flash also available in Google Antigravity and the Gemini Enterprise Agent Platform.
