Ookla vs Fast.com - Which Speed Test Should You Trust?
You ran Speedtest.net and got 310 Mbps. You opened Fast.com right after and got 190 Mbps. Same house, same router, same device, two minutes apart. One of them is lying, right?
Neither one is lying. They're measuring genuinely different things. Once you understand what each tool is actually doing, the gap between their numbers makes a lot of sense. This guide breaks down how each major speed test works, explains why the results differ, and tells you which tool to use depending on what you're trying to figure out.
Why Different Speed Tests Give Different Results
The most common misconception is that all speed tests measure the same thing — how fast is my internet? They don't. Each tool has a different design, different test servers, a different number of parallel connections, and a different idea of what counts as a valid result. Those choices add up to meaningfully different numbers.
Think of it like measuring the speed of a river. You can measure peak flow through the widest part with every channel open. You can measure flow through one specific channel that represents real-world traffic. You can measure flow under normal conditions versus during a flood. All three measurements describe the same river. None of them is wrong. They just answer different questions.
Speed test disagreements fall into a few main categories:
- Server location: A test to a server 5 miles away will return higher throughput than one 500 miles away. Latency and routing overhead grow with distance.
- Number of parallel connections: A test that opens 8 simultaneous TCP streams will push your connection much harder than one that uses a single stream.
- What network they route through: Some tests go through your ISP's peering with a generic CDN. Others route specifically through Netflix's content delivery network or Cloudflare's edge. Those paths may be treated differently by your ISP.
- Test duration: A test that runs for 10 seconds gets past TCP slow-start and sees steadier throughput. A short test may never fully open the TCP window.
Knowing this makes the Ookla vs Fast.com comparison much easier to understand.
How Ookla (Speedtest.net) Works
Ookla's Speedtest.net is the oldest major consumer speed test still in active development. It's been around since 2006 and has gone through several generations. The current version uses what Ookla calls a multi-connection test with an adaptive stream count.
Server Selection
When you open Speedtest.net, it pings a list of nearby test servers — hosted by ISPs, data centers, and Ookla itself — and picks the one with the lowest latency. This is almost always a server very close to you, often run by your own ISP or a major carrier in your area. That proximity is intentional: Ookla is trying to measure your connection's raw capability, not the performance of a specific content provider's delivery network.
Multi-Connection TCP
Ookla opens multiple parallel TCP connections at once — typically between 4 and 16, depending on your connection speed. The reason is straightforward. A single TCP connection on a fast link is limited by the TCP congestion window. Modern home connections in the 200–1000 Mbps range can outrun what a single TCP stream will sustain. Multiple parallel streams push the connection closer to its real capacity.
For a 1 Gbps fiber connection, a single TCP stream might show 400–600 Mbps while 8 parallel streams will push 950+ Mbps. Ookla's method is specifically designed to show you the upper limit of what your connection can deliver.
What Ookla Is Actually Measuring
Ookla measures maximum total throughput to a nearby, well-connected test server. It's answering: how many bits per second can my connection move when pushed hard with multiple streams to a close server? That's a useful number. It tells you whether your ISP is delivering the plan you're paying for. It doesn't tell you whether Netflix will buffer, whether your VoIP calls will drop, or whether a specific provider is being throttled.
Ookla also measures ping (round-trip latency) and jitter to the selected test server, and upload speed using the same multi-connection method in reverse.
How Fast.com Works
Fast.com was built by Netflix and launched in 2016. It's not trying to compete with Ookla. It's answering a fundamentally different question.
Single Netflix CDN Server
Fast.com connects to Netflix's own content delivery network (Open Connect, Netflix's proprietary CDN). It doesn't find the closest generic test server. It connects to whichever Netflix CDN node your ISP normally routes Netflix traffic through. For most users in the US, this is a Netflix server physically installed inside the ISP's own network, or at a nearby internet exchange.
This distinction matters a lot. Netflix has peering agreements and physical hardware placements with major ISPs. How your ISP handles traffic to that Netflix hardware is the whole point of Fast.com.
Single Connection, Simulated Streaming Traffic
Fast.com uses a small number of parallel connections (typically 1–4) and simulates the kind of traffic pattern Netflix actually uses during playback. It's not trying to maximize throughput. It's trying to replicate what happens when you actually hit play on a 4K HDR title.
What Fast.com Is Actually Measuring
Fast.com measures your connection's performance specifically to Netflix's delivery network, using traffic patterns that look like real Netflix playback. It's answering: how fast will Netflix actually load for me right now? If Fast.com says 190 Mbps and Speedtest.net says 310 Mbps, neither is wrong. It means your connection can push 310 Mbps to a nearby test server but only delivers 190 Mbps when talking to the Netflix CDN specifically.
Fast.com doesn't show jitter or a meaningful ping value in its basic interface. It shows download speed and, after clicking a button, upload speed and latency. The interface is kept simple because it was built for non-technical users who just want to know if Netflix is slow.
Side-by-Side Comparison: Ookla vs Fast.com
| Attribute | Ookla Speedtest.net | Fast.com |
|---|---|---|
| Owned by | Ziff Davis (acquired 2014) | Netflix |
| Test server type | Nearest ISP or datacenter node | Netflix Open Connect CDN |
| Parallel connections | 4-16 (adaptive) | 1-4 (low) |
| Primary methodology | Max aggregate TCP throughput | Simulated streaming download |
| Best answers | Is my ISP delivering my plan speed? | Will Netflix buffer on my connection? |
| Shows ping / latency | Yes (idle and loaded) | Yes (basic, click to reveal) |
| Shows upload speed | Yes (default) | Yes (click to reveal) |
| Shows jitter | Yes | No |
| Shows packet loss | No (basic), Yes (CLI) | No |
| Measures bufferbloat | No | No |
| Free tier | Yes (with ads) | Yes (no ads) |
| Paid/premium features | Yes (Ookla for Teams, API access) | No paid tier |
| CLI tool available | Yes (speedtest-cli) | No official CLI |
| Privacy concern | Collects IP, location, ISP, results | Minimal data collection |
The Key Difference: Maximum Throughput vs CDN Performance
Here's what most guides get wrong: they present Ookla as the "real" speed test and Fast.com as an approximation. That framing is backwards. Neither is more real than the other. They're measuring different parts of your network's performance.
Ookla is measuring maximum theoretical throughput to your ISP's closest interconnect. Think of your internet connection like a pipe. Ookla is measuring the pipe's diameter when it's running at full pressure to a faucet right around the corner.
Fast.com is measuring what actually comes out when you open the Netflix faucet — which may be connected to a different part of the plumbing entirely, with different pressure, different flow limits, and potentially a valve your ISP is partly closing.
The practical point: a 100 Mbps gap between Ookla and Fast.com isn't noise or error. It's a signal. It tells you something specific about the path from your ISP to Netflix's infrastructure.
When Ookla Beats Fast.com by a Wide Margin
If Ookla consistently shows a download speed 30% or more higher than Fast.com on the same connection, a few things could explain it:
ISP throttling Netflix traffic: This is the most important scenario. Some ISPs have been documented throttling traffic to specific video streaming providers while leaving general internet performance untouched. Because Ookla uses its own test servers (not Netflix's), the throttling doesn't affect the Ookla result. Fast.com uses Netflix's actual servers, so it hits the throttle directly. The FCC Open Internet era saw several major ISPs caught doing exactly this. Net neutrality enforcement has varied by administration, and the practice hasn't been fully eliminated.
Netflix CDN peering is congested: Your ISP may simply have underprovisioned its connection to Netflix at a specific peering point. This isn't malicious but has similar effects on Fast.com results. The traffic isn't being throttled — it's just competing for bandwidth at a congested handoff point.
Ookla server is hosted by your ISP: Many ISPs host Ookla test servers inside their own network. Traffic to those servers never touches the public internet at all. It stays entirely within the ISP's infrastructure, which means it can show full plan speeds even when the ISP's connection to outside content providers is degraded. This isn't fraud — but it can mislead users into thinking their plan is fine when real-world performance is poor.
Distance difference: If the Ookla server is 5ms away and the Fast.com Netflix CDN node is 40ms away, some of that gap is just geography. Not all of it — but some.
Detecting ISP Throttling: Using Both Tools Together
Running Ookla and Fast.com together and comparing the results is one of the most practical throttling detection techniques available to consumers without specialized equipment.
Here's a step-by-step approach:
- Run Ookla three times at different hours (morning, afternoon, evening). Note the average download speed.
- Run Fast.com three times at the same hours. Note the average.
- Calculate the ratio: Fast.com result divided by Ookla result.
- A ratio of 0.85 or higher (Fast.com is within 15% of Ookla) is generally normal. It reflects routing and methodology differences.
- A ratio below 0.65 (Fast.com is 35% or more slower than Ookla) is worth investigating further — especially if it's consistent across different times of day.
- Add a Cloudflare speed test (speed.cloudflare.com) to get a third data point from a different content provider.
If Fast.com is consistently slow but Cloudflare and Ookla are both fast, Netflix traffic is specifically affected. If Cloudflare is also slow but Ookla is fast, the issue is more general — it's your ISP's outbound performance to the public internet, not just Netflix.
A VPN test can help confirm throttling: if your speeds to Netflix improve a lot when using a VPN, your ISP is likely shaping that traffic. The VPN tunnels the Netflix packets inside encrypted traffic, making them look generic, which bypasses the throttling policy.
Cloudflare Speed Test: The Third Option
speed.cloudflare.com is the most technically sophisticated free consumer speed test available — and it's chronically underused. Cloudflare built it as a demonstration of network measurement capability and to test their own global infrastructure.
Unloaded vs Loaded Latency
Cloudflare's test measures both unloaded latency (ping with no transfer happening) and loaded latency (ping measured while a big download is in progress). The gap between these two numbers tells you something important: whether your connection has bufferbloat.
Bufferbloat happens when your router or modem has a large buffer that fills up during heavy transfers. While the buffer is full, new packets queue behind the existing ones — causing latency to spike. A connection that shows 15ms unloaded but 400ms loaded has severe bufferbloat. It'll feel laggy for gaming, video calls, and anything latency-sensitive whenever someone else on the network is downloading something.
Neither Ookla nor Fast.com measures this. Cloudflare does, and it surfaces it clearly in the results.
Single Connection Architecture
Cloudflare's test uses Cloudflare's own edge nodes, deployed in hundreds of cities globally. It typically hits a node very close to you (often sub-10ms). It measures download speed, upload speed, latency under load, and a quality score that combines these metrics.
What Cloudflare Misses
Cloudflare's test doesn't show jitter the way Ookla does, and it doesn't simulate any specific content provider's traffic the way Fast.com does. It's the best option for general network health diagnostics — especially bufferbloat — but it doesn't replace Fast.com for Netflix-specific testing or Ookla for raw speed benchmarking.
Google Speed Test: Convenient but Limited
Typing "internet speed test" into Google surfaces Google's own embedded speed test, powered by Measurement Lab (M-Lab). It's fine for a quick sanity check and it has the advantage of running directly in Google's search results without going anywhere else.
But it uses a single connection, a single M-Lab server, and a relatively short test duration. The result will typically be lower than Ookla on fast connections because a single TCP stream can't saturate a fast link. M-Lab's infrastructure is also research-focused, not built specifically for consumer-facing throughput testing.
Google's test is useful for a fast yes/no check: is there internet connectivity and is it obviously broken? For any serious diagnosis, use Ookla, Cloudflare, or Fast.com instead.
Result Comparison: Same Connection, Four Tools
The chart below shows what's common when you run multiple speed tests on the same connection in quick succession. The hypothetical connection has a 300 Mbps download plan. The variation across tools isn't error — it reflects the methodology differences discussed throughout this guide.
Notice that Ookla returns the highest result, closest to the plan speed. Fast.com is the lowest because it routes specifically through Netflix CDN infrastructure. Cloudflare is close to Ookla because Cloudflare's edge is well-peered with most ISPs. Google lands in the middle, limited by its single-stream approach.
How Each Tool Routes Your Test Traffic
The diagram below shows a simplified routing map for each tool. Your router is on the left. The three tools take different paths to their measurement endpoints.
The diagram highlights why the same physical connection produces three different measurements. Ookla stays close to home. Fast.com reaches Netflix's infrastructure (which may have different provisioning). Cloudflare hits a geographically close but independently operated edge node.
Six-Tool Comparison: Full Feature Matrix
| Feature | Ookla | Fast.com | Cloudflare | Google / M-Lab | speedtest.now | LibreSpeed |
|---|---|---|---|---|---|---|
| Download test | Yes | Yes | Yes | Yes | Yes | Yes |
| Upload test | Yes | Yes (click) | Yes | Yes | Yes | Yes |
| Ping / latency | Yes (idle + loaded) | Basic | Yes (idle + loaded) | Yes (basic) | Yes (idle + loaded) | Yes |
| Jitter measurement | Yes | No | Yes | No | Yes | Yes |
| Packet loss measurement | No (basic), Yes (CLI) | No | No | No | Yes | No |
| Bufferbloat / loaded latency | No | No | Yes | No | Yes (dedicated test) | No |
| Gaming performance test | No | No | No | No | Yes | No |
| Wi-Fi diagnostics | No | No | No | No | Yes | No |
| ISP comparison data | Yes (Ookla database) | No | No | Partial (M-Lab) | Yes (230+ ISPs, 77 countries) | No |
| Parallel connections | 4–16 | 1–4 | Single + multi | Single | Multi-stream | Configurable |
| Open source | No | No | Partial | Yes (M-Lab) | No | Yes (fully) |
| Ads / commercial | Yes (free tier) | No | No | No | No | No |
A Note on speedtest.now
speedtest.now is our partner platform and the tool we recommend for full diagnostics. It's the only consumer-facing option that tests all five speed test metrics (download, upload, ping, jitter, and packet loss) in a single run, adds a dedicated loaded latency test for bufferbloat detection, a gaming performance test, and a Wi-Fi diagnostics mode. Its ISP comparison database — covering 230+ providers in 77 countries — lets you see how your results compare to others on your carrier. If you want one tool that covers everything on this page, it's the right pick.
A Note on LibreSpeed
LibreSpeed deserves more recognition than it gets. It's a fully open-source speed test you can self-host on any server. Network admins running internal diagnostics use it to test performance to a specific server they control. If you want to test your connection's performance to your company's data center, or measure a specific network segment, LibreSpeed lets you set up a test server exactly where you need it. Most everyday users won't need it — but if you're a homelabber or sysadmin, it belongs in your toolkit.
Scenario Guide: Which Tool to Use When
| Scenario | Recommended Tool | Why |
|---|---|---|
| Check if ISP is delivering my plan speed | Ookla | Multi-stream test to nearby server gives closest approximation to plan capacity |
| Netflix keeps buffering or dropping quality | Fast.com | Tests specifically against Netflix CDN using Netflix-like traffic patterns |
| Suspect ISP is throttling Netflix | Both Ookla and Fast.com | Compare: if Ookla is fast but Fast.com is slow, Netflix traffic is being shaped |
| Gaming lag and high latency during downloads | speedtest.now (gaming test) or Cloudflare | speedtest.now's dedicated gaming test measures ping under load; Cloudflare loaded latency reveals bufferbloat |
| Video call quality problems (Zoom, Teams) | speedtest.now or Ookla + Cloudflare | speedtest.now tests all five metrics in one run including jitter and packet loss; Cloudflare also shows loaded latency |
| Full five-metric diagnostic (all at once) | speedtest.now | Only consumer tool that tests download, upload, ping, jitter, and packet loss in a single run |
| Benchmark against your ISP or country average | speedtest.now or Ookla | speedtest.now covers 230+ ISPs in 77 countries; Ookla has the largest global database |
| Troubleshoot slow internet (unknown cause) | Run all three: Ookla, Fast.com, Cloudflare | Pattern across tools narrows down whether issue is ISP, specific CDN, or equipment |
| Quick sanity check, no app needed | Google Speed Test | Available instantly in search results, adequate for confirming basic connectivity |
| Testing internal network / server performance | LibreSpeed (self-hosted) | Control both endpoints; test any specific network path you need |
| Documenting ISP performance over time | Ookla (with history) or Cloudflare | Both retain result history; Ookla has the largest comparison database |
| Testing whether a VPN is slowing you down | Ookla before and after VPN | Consistent methodology makes before/after comparison clean |
Gaming, Streaming, and Working from Home: Specific Guidance
Gaming
For competitive online gaming, raw download speed matters far less than most players think. A 50 Mbps connection with 8ms ping and no bufferbloat will beat a 500 Mbps connection with 200ms loaded latency every time. Use Ookla to confirm your raw download and upload speeds meet the game's minimum requirements. Then use Cloudflare's loaded latency measurement to check for bufferbloat. If loaded latency is more than 3x your unloaded latency, your router's buffer settings are hurting your game performance — regardless of what your plan speed is.
Streaming (Netflix, YouTube, Disney+)
Netflix needs roughly 25 Mbps for 4K HDR and 5 Mbps for 1080p. The numbers aren't the problem on most modern connections. The question is whether your connection's path to Netflix's CDN is congested or throttled. Fast.com is the right diagnostic tool here. If Fast.com shows 25+ Mbps and Netflix is still buffering, the problem is more likely the Netflix app, your TV's processing, or Wi-Fi signal quality to the specific device — not your internet connection.
Video Calls
Zoom needs about 3 Mbps up and down for HD group calls. Teams and Google Meet are similar. The bandwidth requirement is low. The issue is almost always latency, jitter, or packet loss. A connection that shows 200 Mbps download but 80ms jitter will produce choppy audio and frozen video. Run Ookla and look at the jitter number. If it's above 30ms consistently, that's your problem. Cloudflare's loaded latency test will also reveal whether jitter gets worse under load — which matters if calls degrade when someone else in the house is streaming.
The Honest Answer: Run at Least Two Tools
Any single speed test result is a limited snapshot. It represents one measurement method, one server location, one set of TCP parameters, at one moment in time. Running a single test and concluding your internet is definitively fast or slow is like measuring your blood pressure once and making permanent health decisions from it.
The practical standard for network troubleshooting is to run at least two different tools and compare the results. The combination that gives you the most information with the least effort is Ookla plus Cloudflare. Ookla gives you raw throughput and jitter. Cloudflare gives you latency under load and bufferbloat. If you have a Netflix-specific issue, add Fast.com as a third measurement.
Run each test at least twice to rule out one-time flukes. Run them at different times of day if you suspect congestion — ISP network behavior at 2pm differs from 8pm when residential usage peaks. If you're trying to document an ongoing problem for your ISP's customer service, screenshot the results and note the time. ISPs respond much better to pattern evidence than to one-time complaints.
The goal isn't to find the highest number. The goal is to understand what your connection is actually doing on the paths that matter for how you use it.
Summary: What Each Number Actually Tells You
Ookla's high number tells you your ISP's pipe to nearby infrastructure is healthy. That's worth knowing — but it's not the full story.
Fast.com's number tells you whether Netflix traffic specifically is getting through at adequate speed. If it's low while Ookla is high, you've identified a path-specific performance problem.
Cloudflare's loaded latency tells you how your connection handles simultaneous speed and latency demands — which is the real test for any real-time application.
Google's number is a quick sanity check. Nothing more, nothing less.
Put them together and you have an actual picture of your connection's behavior — not just one number to argue with your ISP about. That's the point.