Google Search said it is now using the latest Gemini model named 3.5 Flash-Lite. It is being used for agentic search experiences, but also likely for Google AI Overviews and AI Mode (I assume). Google wrote, "3.5 Flash-Lite is also rolling out in Google Search," on the company news blog.
Google said where it is being used in Google Search by saying "we’re also releasing Gemini 3.5 Flash-Lite, designed for both low-latency tasks and tasks where high throughput is critical for developers workflows, like agentic search and document processing." Google specifically said "agentic search," as an example.
Google announced information agents and agentic search features at Google I/O in May with Gemini 3.5.
But Google is likely also using 3.5 Flash-Lite as an updated model for Google AI Overviews and AI Mode, if not now, probably soon.
Robby Stein from Google later said on X, "It offers stronger instruction following and better understands user intent, so conversations flow much more seamlessly." Where exactly? In AI Mode? AI Overviews? Just agentic search? I am not 100% sure. Well, the day after, Rajan Patel of Google added, "Yes. It'll be one of the models that Search queries get routed to, depending on the question - excited because we're seeing it's especially great at understanding user intent for conversational q's," when asked about this.
3.5 Flash-Lite is Google's "fastest, most cost-effective 3.5-class model, delivering 350 output tokens per second according to the Artificial Analysis Index, also significantly outperforming prior Flash-Lite generations in agentic workflows."
Google added:
3.5 Flash-Lite enables efficient scaling for agentic systems. Across thinking levels, the model significantly outperforms 3.1 Flash-Lite. Depending on the workload, developers can configure the model to prioritize low-latency, low-cost execution for high-volume tasks with the minimal and low thinking levels, or engage higher thinking levels to process multi-step subagent workloads. The model now also has computer use as a built-in tool to reliably support these agentic tasks across surfaces.It’s a significant step up in coding and agentic tasks as seen in Terminal-Bench 2.1 (54% vs 31%), long context as seen in GDM-MRCR v2 (72.2% vs. 60.1%), and real-world task execution as seen in GDPval-AA v2 (1140 vs. 642).
In fact, on many agentic and coding evals, 3.5 Flash-Lite even outperforms 3 Flash, including on SWE-Bench Pro (54.2% vs. 49.6%) and OSWorld-Verified (74.0% vs. 65.1%), making it a faster & more capable option for workloads on both 2.5 and 3 Flash.
Here are some performance charts Google shared:
So maybe Google Search will get a bit faster as well?
Search information agents are coming. This lays the foundation for that IMO. At I/O, Google said Search agents would launch this summer. Well, it's summer. :)
— Glenn Gabe (@glenngabe) July 21, 2026
From I/O: "Information agents will launch first for Google AI Pro & Ultra subscribers this summer." https://t.co/Jgi6Qrg7IL pic.twitter.com/mpByMSz3xo
Great momentum from @GoogleDeepMind with today’s launch of three new Gemini models.
— Robby Stein (@rmstein) July 21, 2026
Excited to start rolling out Gemini 3.5 Flash-Lite in Search today. It offers stronger instruction following and better understands user intent, so conversations flow much more seamlessly. https://t.co/DFnL6YPVVP
Hey Barry, check out the tweet from @rmstein here: https://t.co/CGHPZvi5Qn
— Rajan Patel (@rajanpatel) July 21, 2026
Yes. It'll be one of the models that Search queries get routed to, depending on the question - excited because we're seeing it's especially great at understanding user intent for conversational q's
— Rajan Patel (@rajanpatel) July 22, 2026
Forum discussion at X.
