Your bandwidth.
All of it.
Symmetric service, dedicated optical wavelengths and a 1:1 load factor. Native IPv6 and BGP support for the networks you operate.
Explore Dedicated InternetAI connectivity / New YorkSuper Intelligence.
Super Intelligence.
Connected.
Intelligence moves through infrastructure. Dedicated Internet from Stealth supports AI applications, agents and real time interactions.
Our target: 10 ms or less for most covered endpoints. These are network round trips from Stealth NYC to public AI service endpoints; they do not measure AI inference or response time. See endpoint targets and the AI connectivity SLA for service conditions.
01 / The live connectionEvery round trip
Every round trip
has a story.
A rolling five days of ICMP round trip time to chat websites and API endpoints. Five minute averages, in milliseconds.
Vantage pointStealth NYC32 Avenue of the Americas
Current ICMP reachability12 / 12Endpoints responding to fresh ICMP checks
Observation windowFive daysSuccessful recordings; empty periods remain gaps
Measurements available. The display refreshes every 60 seconds.
Network investigations · Observations, assessments and evidence gaps from the latest 14 days.
ICMP RTT / Milliseconds
Rolling five days · UTC
Enter time range
Five-minute bucket times use the selected timezone and include the UTC offset.
Show
Solid lines: API. Dashed lines: chat website. Hover or tap to inspect; drag across the chart to zoom. On touchscreens, first choose Select time range. Use ← / → when focused to inspect a bucket. Escape or Reset zoom returns to the rolling view. Hollow markers indicate a partial bucket. The vertical scale adjusts to the selected endpoints and time range.
Select a point to inspect its five minute average. No samples are plotted before collection began.
Find the right connection for your building.
Check availabilityThe latest observations.
RTT is the latest successful recording. ICMP and TCP status come from separate current probe and scrape checks.
The table’s p95 uses available recordings across up to five days. The SLA assesses p95 separately in five-minute intervals; a reading alone does not establish an SLA breach.
| Public endpoint | Surface | Latest RTT | Median | p95 | ICMP | TCP :443 | RTT recorded / UTC |
|---|---|---|---|---|---|---|---|
| OpenAIchatgpt.com | Chat website | 1.585 | 1.375 | 10.006 | Responding | Responding | 26 seconds ago |
| OpenAIapi.openai.com | API | 1.449 | 1.418 | 1.874 | Responding | Responding | 26 seconds ago |
| Anthropicclaude.ai | Chat website | 1.656 | 1.469 | 1.604 | Responding | Responding | 26 seconds ago |
| Anthropicapi.anthropic.com | API | 1.467 | 1.479 | 1.615 | Responding | Responding | 26 seconds ago |
| Googlegemini.google.com | Chat website | 0.711 | 0.678 | 0.758 | Responding | Responding | 26 seconds ago |
| Googlegenerativelanguage.googleapis.com | API | 0.423 | 0.440 | 0.698 | Responding | Responding | 26 seconds ago |
| Microsoftcopilot.com | Chat website | 6.768 | 6.717 | 22.868 | Responding | Responding | 26 seconds ago |
| Microsoftcopilot.microsoft.com | Chat website | 1.005 | 1.085 | 1.629 | Responding | Responding | 26 seconds ago |
| xAIapi.x.ai | API | 1.472 | 1.326 | 1.479 | Responding | Responding | 26 seconds ago |
| DeepSeekapi.deepseek.com | API | 0.711 | 0.675 | 1.135 | Responding | Responding | 26 seconds ago |
| Perplexityapi.perplexity.ai | API | 1.064 | 1.276 | 1.446 | Responding | Responding | 26 seconds ago |
| Mistralapi.mistral.ai | API | 1.203 | 1.337 | 1.803 | Responding | Responding | 26 seconds ago |
Snapshot retrieved 09 Nov 2026 21:21:40 UTC. A fresh page or chart update is not evidence of a fresh RTT measurement.
Explore chart data as a table
Load the chart to browse every five minute bucket, including gaps.
| Bucket ending / UTC | RTT / ms | Successful samples | Coverage |
|---|---|---|---|
| Waiting for chart data. | |||
Dedicated Internet / AI connectivity SLAA network commitment.
A network commitment.
With clear remedies.
Explore the AI connectivity SLA for named API endpoints and chat websites, measured from Stealth NYC.
Defined latency targets, a 10% service credit for a qualifying Stealth breach, and a cancellation right after five affected business days within 30 days. Customer notice and the applicable signed agreement govern eligibility.
Read the AI connectivity SLA02 / Infrastructure for the interactionYour workload.
Your workload.
Our network.
Your bandwidth.
A path to your
cloud workloads.
Connect locations and cloud environments with Stealth’s Cloud Connect and private Ethernet services. Confirm providers, regions and connection options with our team.
Explore Cloud ConnectBuilt here.
Operated here.
Our fiber. Our construction crews. Our network operations. Infrastructure for the applications, agents and conversations your work depends on.
Check your building For property owners & managersA view of the network.
These are IPv4 ICMP round trip measurements from Stealth NYC at 32 Avenue of the Americas to public service endpoints. The endpoints may resolve to CDN edges. These measurements do not measure model inference time, API response time or the performance of a particular AI model.
The chart shows five minute arithmetic averages of successful recordings collected every 30 seconds. Median and p95 use the underlying successful recordings in the available window of five days. Missing periods remain gaps. A hollow point marks fewer than ten successful samples or an unavailable sample count.
ICMP and TCP port 443 status are checked independently using probe success and collection health. A TCP connection does not establish application availability. Status older than two minutes is unconfirmed; an older successful RTT does not mean a target is responding now.
Plan the connectionWhat are
What are
you building?
Tell us about your AI workloads, location and bandwidth requirements. We’ll help you plan the connection.
Discuss AI connectivity