A new generation of Cortex Discover combines faster browser startup, responsive interactions, and efficient resource use with an integrated agentic AI experience.
A browser should not become slower simply because it becomes smarter.
That principle has shaped our performance work on Cortex Discover from the beginning. As AI moves deeper into everyday browsing, research, development, and enterprise workflows, performance becomes more than a benchmark. It becomes part of the user experience.
In our previous performance study, we explored how Cortex Discover could maintain browser responsiveness while supporting persistent AI capabilities. We also established a more rigorous approach to measuring resource efficiency, moving beyond isolated benchmark results toward controlled, repeatable testing.
With Cortex Discover 1.5, we are taking that work forward.
The latest version demonstrates faster startup, near-identical memory consumption compared with Chrome, competitive rendering performance, and encouraging improvements in interaction responsiveness and media processing.
The goal remains simple: more intelligence, without unnecessary browser overhead.
A Faster Start to Every Session
The first interaction with a browser happens before a page even loads.
Whether users are opening technical documentation, returning to an active project, or beginning a new research session, startup speed shapes how quickly they can get to work.
In controlled testing against Chrome 156, Cortex Discover 1.5 delivered:
- 14.8% faster native startup, completing in 1,039 ms compared with Chrome’s 1,220 ms.
- 6.1% faster warm startup, completing in 966 ms compared with Chrome’s 1,028 ms.
The native startup result also extends the startup advantage seen in our earlier performance work.
For users, the value is straightforward. A faster browser removes unnecessary friction between opening an application and getting work done.
For an agentic browser, that responsiveness matters even more. AI capabilities should expand what users can accomplish, not slow down the experience before they begin.
More Intelligence. Nearly Identical Memory.
One of the central challenges in building an agentic browser is balancing capability with resource consumption.
Traditional browsers already manage pages, rendering, media, extensions, and increasingly complex web applications. An agentic browser adds another dimension: AI that can understand context, assist with workflows, and remain available as users move between tasks.
That intelligence must coexist with the browser’s fundamental responsibilities.
Our latest tests show Cortex Discover 1.5 maintaining near memory parity with Chrome across multiple workloads.
Across single tab, multi tab and active workloads, Cortex Discover stayed effectively at memory parity with Chrome, demonstrating that integrated AI does not have to come with significantly higher everyday browser memory use.
Single-tab memory was approximately 0.4% lower in Cortex Discover, while multi-tab and active workload measurements remained effectively equivalent.
These results demonstrate that adding an integrated AI experience does not automatically require significantly higher baseline memory consumption during ordinary browsing.
That distinction becomes increasingly important as AI moves from an occasional assistant into a persistent part of the digital workspace.
Responsiveness Beyond Startup
Fast startup is valuable, but a browser also needs to stay responsive once users begin interacting with it.
Switching between pages, navigating documentation, interacting with web applications, and returning to active research all contribute to the perceived speed of a browser.
Cortex Discover 1.5 maintained strong interaction performance in the latest controlled comparison.
Browser-wide interaction-tail p95 measured 52.01 ms for Cortex Discover compared with 54.01 ms for Chrome, representing a 3.7% faster point estimate.
Background readiness also showed a favorable result, measuring 68.08 ms compared with 91.02 ms, a 25.2% faster point estimate in that workload.
A separate controlled interaction workload covering 5,000 interactions per browser showed effectively identical handler-to-first-frame latency.
The takeaway is not that every interaction becomes faster. It is that Cortex Discover can add deeper AI capabilities while preserving the responsive browser experience users expect.
That becomes increasingly important as the browser evolves from a window into the web into an active environment for intelligent work.
Full-Speed Rendering, Efficient Media
Modern browsing extends far beyond static webpages.
Users interact with dashboards, video, visual applications, development environments, documentation portals, and increasingly sophisticated web interfaces.
Cortex Discover 1.5 maintained 60 FPS rendering, matching Chrome in the controlled workload. Fixture throughput was also effectively equivalent, with Cortex Discover measuring 60.51 compared with Chrome’s 60.63.
Media performance showed additional encouraging results. In the broader video workload, Cortex Discover consumed 1.4% less memory, while both browsers recorded zero dropped frames.
A separate compact deterministic H.264 test measured:
- 12.8% lower video CPU usage
- 12.9% less encoded CPU time
- 0.8% lower memory consumption
- Nearly identical decoded output at approximately 30 FPS
These measurements come from defined controlled workloads rather than a universal media or battery-life benchmark.
But they illustrate an important direction for Cortex Discover: improving processing efficiency while preserving the visual performance users expect from a modern browser.
Keeping AI Close Without Interrupting the Browser
One of Cortex Discover’s defining capabilities is the integration of Cortex directly into the browsing experience.
Instead of repeatedly moving between separate applications, users can access AI assistance while working with pages, documentation, research, and connected workflows.
That experience depends on the Cortex interface remaining responsive and predictable.
In the latest short-duration lifecycle measurements, opening the Cortex panel created no additional panel process. Open-idle CPU increased by only 0.224 percentage points of one core.
Panel render readiness measured approximately 26 ms at p95, while the measured runs retained no additional processes.
The tests also confirmed correct rendering, healthy interface reuse, and network inactivity after settling.
These results support an important part of the Cortex Discover design: keep intelligence close to the user without turning the AI experience into unnecessary background overhead.
Measuring Performance Under Controlled Conditions
Performance comparisons are most useful when they are repeatable and grounded in clearly defined workloads.
For Cortex Discover 1.5, our evaluation covered startup, memory, interaction responsiveness, rendering, video workloads, and native panel behavior.
The testing used repeated runs, fresh browser profiles, workload-completion checks, and consistent conditions intended to reduce measurement noise and make results easier to compare over time. The current measurements represent the systems and workloads tested and are not intended as universal performance claims across every device or operating environment.
Our objective is not to reduce browser performance to a single benchmark number.
It is to establish a repeatable engineering baseline that helps us understand where Cortex Discover is improving and where we should continue investing.
Performance Is Part of the AI Experience
Cortex Discover is designed around a larger idea: the browser is becoming an active environment for intelligent work.
Users increasingly expect AI to understand what they are viewing, connect information across tasks, assist with technical workflows, and help move from questions to actions.
Delivering that experience requires more than adding another AI interface. The browser itself must remain fast, predictable, and efficient.
Startup efficiency reduces the time needed to enter a workflow. Memory efficiency allows the browser to coexist with IDEs and other demanding applications. Responsive interaction makes AI feel like a natural extension of browsing. Disciplined resource management helps prevent background intelligence from becoming unnecessary overhead.
These are not isolated optimizations. Together, they shape how an agentic browser feels during everyday use.
Faster Browser. More Intelligent Workspace.
Cortex Discover 1.5 represents another step toward bringing high-performance browsing and integrated AI into one experience.
The latest controlled results demonstrate meaningful startup improvements, near memory parity with Chrome, competitive interaction responsiveness, equivalent rendering performance, and encouraging media-processing efficiency.
Just as importantly, they reinforce the engineering direction established in our earlier research:
Intelligence and browser performance should advance together.
As Cortex Discover evolves, we will continue improving responsiveness, efficiency, and the experience of working with AI directly where users browse, research, develop, and make decisions. Because the future of browsing is not simply about making AI available inside a browser.
It is about making intelligence feel like a natural part of a fast, connected, and productive workspace.
Cortex Discover 1.5. Faster Starts. Smarter Browsing. Intelligence in Motion.
