ESX, QBCore or Qbox in 2026: which FiveM framework to pick
ESX, QBCore and Qbox compared in October 2026: which one to pick, what txAdmin offers, migrations and whether to build your own.
The framework is the first decision on a FiveM server and the hardest one to undo. It decides which scripts you can buy, what your database looks like, who you can hire, and how much a migration will cost you in two years.
I have run an ESX server for years, and every XeX script supports ESX, QBCore and Qbox, so I spend a lot of time inside all three. This is where they stand in October 2026, checked against their repositories, and what I would pick in each situation.
The short answer
| Your situation | What I would pick |
|---|---|
| New roleplay server, starting from zero | Qbox, if you can run MariaDB. ESX if your team already knows it. |
| Existing ESX server | Stay on ESX and keep it updated. |
| Existing QBCore server | Plan the move to Qbox for a quiet month. |
| Racing, drift, freeroam or a custom game mode | No framework, or your own small core. |
The rest of this post is the reasoning.
Where you start: txAdmin’s recipes
Cfx.re does not have an official roleplay framework. What it does have is txAdmin, which ships with FXServer, and its setup wizard offers ready-made recipes. These are the FiveM ones on its official list, as of October 2026:
| Recipe | What it sets up |
|---|---|
| FiveM Basic Server | A minimal server, no framework |
| ESX Legacy | The most used roleplay framework, with jobs, housing and vehicles |
| Qbox | A modern framework compatible with QBCore resources |
| QBCore | A roleplay framework with jobs, gangs and housing |
| vMenu | A server menu with permissions, no roleplay |
| StreetKings | A racing game mode with missions and events |
| WarfareTacticsV | A real-time strategy game mode |
Each recipe is maintained by its own project. That list is a good picture of the market: almost every roleplay server starts with one of those three frameworks.
The three, side by side
| ESX Legacy | QBCore | Qbox | |
|---|---|---|---|
| Latest version (Oct 2026) | 1.16.2, 17 Oct 2026 | 1.3.0, no published releases | 1.24.0, Aug 2026 |
| Pace | Several releases a month | Slow | Long gaps, big releases |
| Core library | esx_lib (new in 2026) and ox_lib |
Its own qb- resources |
ox_lib |
| Inventory | Its own (new in 1.16), switches to ox_inventory if present | qb-inventory |
ox_inventory, required |
| Database | MySQL or MariaDB via oxmysql | MySQL or MariaDB via oxmysql | MariaDB 10.9+ only |
| OneSync | Supported | Supported | Required |
| Runs QBCore scripts | No | Yes | Most of them, through a bridge |
| Script catalogue | The largest | Large | QBCore’s, plus Qbox-native |
| License | GPL-3.0 | GPL-3.0 | GPL-3.0 |
ESX Legacy
The oldest framework still in active development, and the one with the most scripts written for it. In 2026 it has
moved fast: version 1.16 added its own slot inventory and a Discord whitelist, and es_extended now depends on a new
library, esx_lib. It still switches to ox_inventory automatically when ox_inventory is running, which is how most
large ESX servers run.
The upside is choice. Whatever system you need, someone has written it for ESX, and most developers you could hire know it. The downside is the same catalogue: years of scripts written for ESX v1 still circulate, with patterns that cost you later, like per-player lookups, synchronous database calls and loops every frame. And the 2026 changes mean two stacks now live side by side on many servers, ESX’s own and ox. Read the release notes before updating a server with players on it.
QBCore
QBCore took a large part of the market from 2021 with its own ecosystem of qb- resources. The core moved to the
qbcore-fivem organization on GitHub (old links redirect) and has slowed down: qb-core sits at 1.3.0 without
published releases, while most of the modern ox stack now targets Qbox instead. ox_inventory’s support for QBCore is
labeled experimental, so most QBCore servers stay on qb-inventory.
It is still a fine base with a large catalogue and an API many developers know. But if you are on QBCore today, the question is not whether to stay forever, it is when to move to Qbox.
Qbox
Qbox started in 2022 as a QBCore fork and is now its own framework, built on ox_lib, ox_inventory and ox_target, with a bridge that makes most QBCore scripts work without changes. Version 1.24 in August 2026 focused on security, with SQL injection fixes in search filters and opt-in for admin commands.
It is the most modern of the three, and the strictest. MariaDB 10.9 or newer is required (MySQL is not supported), OneSync and ox_inventory are not optional, and multiple jobs per character cannot be turned off. The bridge covers the documented QBCore API; scripts that read QBCore’s database tables directly, or use its internal files, will not work.
Moving from QBCore is documented step by step: run Qbox’s SQL changes, create the groups table, convert the inventory to ox_inventory, change job grades from text to numbers and run the conversion command. It is a data migration, not a rewrite. Back up first.
Moving from ESX to either QB-based framework has no official tool. It means migrating characters, money, inventories and vehicles, and replacing every ESX-only script. Unless you are rebuilding the server anyway, it rarely pays off.
The ox stack, and ox_core
Whatever framework you pick, you will probably run the same libraries: oxmysql for the database, ox_lib for callbacks, zones, menus and the player cache, and ox_inventory and ox_target. They come from Overextended. In 2025 the original repositories were archived and the community kept them alive; in 2026 development moved back to Overextended, with releases this month. If an old forum thread tells you “ox is dead”, it is out of date.
The same team makes ox_core. It is not a roleplay framework you install and open the next day: it is a modern core for developers, written in TypeScript and usable from Lua, built around OneSync and state bags, with no compatibility layer for ESX or QBCore scripts. It makes sense for a team that will write most of its server itself.
Others you will run into
- vRP is one of the oldest frameworks. Its last tagged release is from 2018, and its own README says vRP 2 will probably be the last major version. In Brazil, many servers run commercial bases derived from vRP, often called Creative; they do not use the ESX or QBCore API, so scripts made for those do not run on them without adapting.
- EssentialMode and ESX v1 belong to another era. If a guide tells you to install EssentialMode, close it.
- No framework at all is the right call for racing, drift or freeroam servers. txAdmin’s basic recipe, ox_lib and ACE permissions cover what they need, without jobs and an economy they will never use.
Should you build your own?
Some servers write their own core. If you are considering it, this is the minimum it has to do:
- Identify the player when they connect, with a stable identifier (
GetPlayerIdentifierByType(source, 'license')), and check bans and the whitelist before letting them in. - Store accounts and characters, with money as integers, using parameterized queries only.
- Give permissions through ACE (
add_principal,add_ace,IsPlayerAceAllowed). - Share data with state bags, and turn on
sv_stateBagStrictModeor validate every change on the server. - Treat every network event as untrusted: take the player from
source, never from the payload. - Expose a stable API with exports and lifecycle events: player loaded, unloaded, job changed.
- Write a bridge for the ESX or QBCore API that the scripts you buy call, or adapt every script yourself.
For a roleplay server, my answer is no. Every script you buy needs a bridge or a rewrite, bridges are a moving target (ESX changed its library and its inventory in the last two months alone), you become the only person finding your security bugs, and nobody on a forum can help you with your framework. Qbox and ESX already give you a base that works. Spend that time on the systems that make your server different.
For a game mode that is not roleplay, or a team with developers to spare, a small core of your own on ox_lib and oxmysql can be the right call. Keep in mind that ox_lib and oxmysql are LGPL-3.0 and ox_inventory is GPL-3.0.
Whatever you pick
Buy scripts with an open bridge. A script that only talks to the framework through a bridge you can read survives the framework’s next big release, and comes with you if you migrate. That is why every XeX script ships its bridge unencrypted, for ESX, QBCore and Qbox.