Every framework comparison on page one of Google was written by someone selling hosting, and not one of them contains a number. We run the crawler that polls every public FiveM server every five minutes. Here is what the fleet actually runs.
Every framework comparison on the first page of Google was written by someone selling hosting or scripts, and not one of them contains a single number. That is the whole problem with this question: it gets answered with vibes. ViceHub polls every public server on the Cfx.re directory every five minutes, so we can answer it with the fleet itself. Here is what 32,616 live FiveM servers were actually running when we last looked.
What the live server list declares
At our last poll — 13 July 2026, 22:05 UTC — 32,616 public FiveM servers were online. Classifying each one by the framework tags its operator set on the listing:
ESX — 11,836 servers
QBCore — 4,640 servers
Qbox — 3,523 servers
vRP — 1,917 servers
More than one framework declared — 299 servers
No framework declared at all — 10,401 servers
So ESX leads QBCore by roughly 2.5 to 1 on the live fleet. That alone contradicts most of what gets written about this, which tends to present QBCore as the default modern choice and ESX as legacy. On the actual server list, ESX is not legacy. It is the incumbent, and it is not close.
The obvious caveat, which the vendor blogs never state: these tags are self-declared. An operator types them into their server configuration. They can be wrong, stale, aspirational, or absent. A server running qbx_core that still tags itself "qbcore" is counted here as QBCore. Nothing forces anyone to be honest or current. What this data is, precisely, is — which is a strong signal, and is still 32,616 more data points than any other comparison you will read.
The median FiveM server peaked at two players. Not two hundred — two. We measured 31,334 of them. Here is what a server realistically needs to feel alive, and why your slot count is fiction.
Everyone tells you MLO stands for Map Loading Object. There is no primary source for that, and it does not matter. Here is what an MLO actually is, and how to install one without breaking your server.
what 32,616 operators chose to say about their own servers
Servers are the wrong unit. Count players.
A framework with 4,000 empty servers is not more popular than one with 1,000 full ones. So instead of counting servers, count the people on them.
For every server we take its daily peak — the highest concurrent player count we observed on it that day — and add those peaks up by framework. Using 12 July 2026, a full day of five-minute polling:
ESX — 124,366 combined daily peak, across 6,845 servers
No framework declared — 118,931, across 5,992 servers
vRP — 55,746, across 1,312 servers
QBCore — 23,167, across 2,558 servers
Qbox — 15,507, across 1,987 servers
Weighted by players, ESX leads QBCore by about 5.4 to 1 — more than double its lead by server count. QBCore's servers are not just fewer, they are smaller. The mean QBCore server peaked at 9.1 players that day. The mean ESX server peaked at 18.2.
(One honest note on that arithmetic: each server's peak happens at its own moment, so these totals are a sum of individual peaks, not a count of people simultaneously online across the network. It is the right way to size relative demand between frameworks; it is not the network's concurrent player count.)
QBCore has no mega-servers
This is the finding that surprised me most. Look at the largest single server in each group on 12 July:
ESX — one server peaked at 1,981 players
No framework declared — 1,952
vRP — 1,856
Qbox — 725
QBCore — 534
The biggest QBCore server on the public directory peaked at barely a quarter of the biggest ESX server. There is no QBCore server in the top tier of FiveM at all.
The reason is mostly history. FiveM's giants are old — the Brazilian vRP cities and the big French ESX communities have been running for years, on stacks they have spent those years hardening. Sort the live directory by player count and the top of the list is dominated by Brazilian vrp servers and French ESX servers running 2,048 slots. QBCore arrived after those communities were already enormous, and it has won an enormous number of new servers without yet displacing any of the incumbents at the top. If you want to inspect this yourself, the live server list is sorted by real player counts and shows each server's population history.
Qbox is where people are building, not where they are playing
Qbox looks like a serious contender on the server count: 3,523 declared servers, about 76% of QBCore's total. That is far more adoption than its reputation as the "new fork" suggests.
Then you look at the players. The median Qbox server peaked at 1 player on 12 July. Its mean peak was 7.8 — the lowest of any framework we measured. Its largest server managed 725.
That is a very specific signature, and it is not a bad one: it is what a framework looks like when developers are actively building on it and players have not arrived yet. Thousands of one-player Qbox servers are thousands of developers with a test server open. Whether that converts into populated communities is the open question for Qbox, and it is genuinely too early to call. What you should not do is read "3,523 servers" as "3,523 thriving communities" — for any framework.
The 10,401 servers that declare nothing
Just under a third of the live fleet — 10,401 servers — set no framework tag at all, and they are not junk. They carried a combined daily peak of 118,931, second only to ESX and more than QBCore and Qbox put together.
Two things are going on. Some of these are the abandoned and half-built servers that make up so much of the directory. But the group also contains a lot of the biggest, most established communities, and that makes sense: a server with 1,900 players does not need to advertise its framework to attract anyone, and many of the largest have drifted so far from stock ESX or QBCore that naming a framework would be misleading anyway. Once a community is big enough, it is running its own thing.
What this data cannot tell you
I would rather you trust the numbers above than be impressed by them, so here is exactly where they are soft:
The tags are self-declared. No one audits them. This measures what operators say they run.
Nothing forces a server to tag at all, and a third of them do not.
Peaks are sampled every five minutes. A spike between two polls is invisible to us.
Only the public directory is counted. Private and unlisted servers do not appear.
A daily peak is not an average. A server that touches 40 players for ten minutes and sits at 3 the rest of the day still records a peak of 40.
None of that changes the shape of the result. ESX leading QBCore 2.5:1 by servers and 5.4:1 by players is far too large a gap to be an artefact of sloppy tagging.
So what should you actually build on?
The data says what the fleet runs. It does not, on its own, say what you should run — a server that peaked at 2 players is a data point about popularity, not a recommendation. So, plainly:
Pick the framework your scripts already support. This is the whole decision and almost everyone overthinks it. Go and find the ten resources you actually intend to run — your inventory, your jobs, your dispatch, your housing — and check what they target. If eight of them are ESX, you are building an ESX server, and no amount of reasoning about code quality changes that. Retrofitting a resource to a framework it was not written for is real engineering work, and you will be doing it ten times before launch. You can see which frameworks the scripts you want actually support on the FiveM scripts marketplace, where every listing declares its framework.
Pick the framework your developer already knows. The most common reason a new server never launches is not that it chose the wrong framework — it is that it never got finished. A developer who is fluent in ESX will ship your server. The same person learning Qbox from scratch may not.
ESX is the safe default in 2026, on the evidence above: the largest script catalogue, the largest player base, and the only framework other than vRP with servers at the very top of the directory. Its reputation for being dated is doing it a disservice on the live fleet.
QBCore is the reasonable choice if you want more built in out of the box and you are building a mid-size community. Just go in knowing the ceiling the data shows: nobody has yet run a thousand-player server on it.
Qbox is the choice if you have a competent developer, you care about code quality, and you are comfortable being early. It is a real project with real adoption. It is not yet a proven home for a large population.
And whichever you choose, choose once. Migrating a live community between frameworks is a rebuild, not an upgrade — every script that touches the framework's exports has to be reworked, and player data has to be moved. That is the actual cost of getting this decision wrong, and it is why it is worth an afternoon of research rather than an hour.
The live server list is updated every five minutes and shows the population history behind every number here, if you want to check the working. Our full ranking and data methodology explains how the crawl works.
Frequently asked questions
Is ESX or QBCore more popular in 2026?
ESX, by a wide margin. Of the 32,616 public FiveM servers online at our last poll on 13 July 2026, 11,836 declared an ESX tag and 4,640 declared a QBCore tag — about 2.5 to 1. Weighted by players rather than servers the gap is wider still, roughly 5.4 to 1. These tags are self-declared by operators, so treat them as a strong signal rather than a census.
Is Qbox actually used, or is it just hype?
It is genuinely used, but mostly by builders rather than players. 3,523 online servers declare a Qbox tag — about 76% of QBCore's server count — yet the median Qbox server peaked at just 1 player on 12 July 2026, and the largest peaked at 725. Qbox has real adoption in development and test servers; it has not yet produced a mega-server.
How many FiveM servers do not declare a framework?
10,401 of the 32,616 online servers — just under a third — declare no framework tag at all. They are not a rounding error: they held the second-largest combined daily peak of any group we measured, which is what you would expect if many of the biggest communities run heavily customised or in-house frameworks and do not advertise it.
Can I migrate from ESX to QBCore or Qbox later?
You can, but treat it as a rebuild rather than an upgrade. Player data, jobs, and every script that touches the framework's exports have to be reworked or bridged. It is far cheaper to pick deliberately at the start than to migrate a live community later.
Which framework should a brand-new server use?
Pick the framework the scripts you actually want already support, and that your developer already knows. That is a more reliable predictor of whether your server ships than any feature comparison — all three frameworks are free, open source, and capable of running a good server.
Half the performance advice in FiveM is wrong, including the famous one about Wait(0) — the official docs mandate exactly what the community tells you to avoid. Here is how to find the real culprit.
