Gunicorn
Pre-fork WSGI server for Python, started with a single command. Our application servers ranking puts Gunicorn at #8 of 14.
On this site vendors can buy visibility, and we say so. Outbound links, including the one to Gunicorn, are marked identically. Read more.
Visit Gunicorn See it in the application servers ranking
Gunicorn at a glance
| Headquarters | Not stated by the vendor |
|---|---|
| Deployment | Vendor does not say, as far as we read |
| Pricing model | MIT-licensed package; the home page lists no prices or plans |
| API | Unknown from our sources |
| Free plan | No free tier on record |
| Best for | Django, Flask or FastAPI teams on UNIX hosts deploying behind Nginx or systemd |
What is Gunicorn?
Gunicorn, short for Green Unicorn, is a Python WSGI HTTP server for UNIX. Its home page shows the whole idea in one line: you point the command at a module and application object, for example gunicorn myapp:app, and the server starts. It uses a pre-fork worker model, in which a master process manages a set of worker processes, and the site calls it battle-tested and production-ready.
The compatible frameworks named on the home page are Django, Flask, FastAPI, Pyramid, Starlette, Falcon, Bottle and Quart. The software is published under the MIT License on a GitHub repository that belongs to the personal account benoitc. The pages we read name no company, so there is no vendor entity, price list or country of establishment to report.
Gunicorn key features
- Pre-fork worker model. A master process starts and supervises separate worker processes that handle requests. This is the design the home page puts forward, and it is the reason the worker count is one of the first settings to tune.
- Several worker types. The default worker is sync. The home page also lists async workers based on Gevent or Eventlet, a threaded worker, and an ASGI worker using native asyncio, so one server can cover different application styles.
- Framework compatibility. Django, Flask, FastAPI, Pyramid, Starlette, Falcon, Bottle and Quart are all named on the home page. The list mixes classic WSGI frameworks with async ones, so teams rarely need a different server per project.
- Layered configuration. Settings can come from environment variables, a gunicorn.conf.py file, the GUNICORN_CMD_ARGS variable or command-line arguments. When the same setting appears twice, the command line takes precedence over the other sources.
- Config inspection commands. The configure page documents gunicorn --print-config to show the resolved settings and gunicorn --check-config to validate them. Both help when a deployment behaves differently from what the config file suggests.
- Deployment documentation. The home page lists documentation topics for running behind Nginx, under systemd and in Docker, plus pages on HTTP/2, dirty arbiters and signals. The pages we read do not describe what dirty arbiters do in detail.
- Optional extras. Extras add the gevent, gthread and tornado worker types, and setproctitle, which makes process names visible in ps and top. Installing only the extras you need keeps the base package small.
Who Gunicorn suits
- Django or Flask teams. A team shipping a conventional WSGI application can start the app with a single command, then move the same settings into gunicorn.conf.py once the deployment settles behind a reverse proxy.
- Operators of async Python services. Teams running FastAPI or Starlette can use the asyncio-based ASGI worker, keeping Gunicorn's process supervision while the framework handles its own concurrency.
- Container and Linux packagers. The install page lists a Docker image on ghcr.io and system packages for Debian, Ubuntu, Fedora and Arch Linux, which suits teams that prefer a distribution package or a ready image over pip.
Getting started with Gunicorn
The install page names pip install gunicorn as the main route, with pipx, a Docker image at ghcr.io/benoitc/gunicorn and system packages for Debian, Ubuntu, Fedora and Arch Linux as alternatives. The package requires Python 3.12 or newer, so older interpreters need a different server or an upgrade first.
After installing, run gunicorn with the module and application name, such as gunicorn myapp:app. For anything beyond a first test, the configure page shows a gunicorn.conf.py example that binds to 127.0.0.1:8000 and sets the worker count to twice the CPU count plus one. Run gunicorn --check-config before the first real deployment to confirm which values are actually in force.
Verdict: Gunicorn
Gunicorn runs Python WSGI applications through a pre-fork worker model, with sync, threaded, Gevent, Eventlet and native asyncio worker types. It requires Python 3.12 or newer and is MIT-licensed.
The GitHub repository belongs to a personal account, benoitc, and no company is named on the pages we read, so there is no vendor contract, support desk or country to point to.
In our application servers ranking Gunicorn stands at #8 of 14: WildFly sits directly above it and Puma directly below. After the challenger places, Gunicorn is one of the established names, ranked in order and without a score.
The vendor states no country of establishment, so Gunicorn is left out of the regional count for this list.
Where Gunicorn wins and where it does not
- Gunicorn runs Python WSGI applications through a pre-fork worker model, with sync, threaded, Gevent, Eventlet and native asyncio worker types.
- Gunicorn is tagged WSGI in our records.
- Catalogue tag: Pre-fork workers.
- Strongest for: Django, Flask or FastAPI teams on UNIX hosts deploying behind Nginx or systemd.
- MIT licence is among the traits we single out for it.
- Python 3.12 or newer is required, which rules out older interpreters on legacy hosts.
- The pages read name no company, so support is whatever the community and repository owner provide.
- Not for: Windows servers or teams wanting a commercial support contract.
Gunicorn pros and cons in practice
- A single command starts an application, so a first deployment takes minutes rather than a configuration project.
- Worker types cover sync, threaded, Gevent, Eventlet and asyncio styles inside one tool.
- The MIT licence and the pip install route make it easy to adopt without procurement.
- Configuration can be inspected with --print-config and --check-config before it reaches production.
- The home page describes a server for UNIX, so Windows hosts are not covered by what we read.
- Python 3.12 or newer is required, which rules out older runtime versions.
- No company or country is named, so there is no vendor contract, support desk or service level to rely on.
- The Uvicorn site calls its Gunicorn worker module deprecated, so teams pairing the two servers need to watch package versions.
What Gunicorn costs
Pricing as recorded: MIT-licensed package; the home page lists no prices or plans.
The home page lists no prices or plans, and the pages we read describe no paid tier, support contract or hosted version. The software is released under the MIT License according to the GitHub repository, which permits use without a licence fee. Running it costs whatever your own servers or containers cost. Support is not described on the pages we read as a service from a company, since the repository belongs to a personal account and no organisation is named. Teams that need a contractual support relationship, an invoice or a named vendor therefore have no route to one on these pages and would need to arrange it through an outside provider or in-house expertise.
Checked against gunicorn.org for Best Application Servers in 2026, review date 15 November 2026.
Gunicorn in practice
Settings come from environment variables, gunicorn.conf.py, GUNICORN_CMD_ARGS and the command line, with the command line winning; gunicorn --print-config and --check-config show and validate the result.
The example config binds to 127.0.0.1:8000 and sets workers to twice the CPU count plus one.
Install options are pip, pipx, a Docker image at ghcr.io/benoitc/gunicorn and Debian, Ubuntu, Fedora and Arch packages.
Documentation covers Nginx, systemd and Docker deployment.
Gunicorn: data and security
Gunicorn is a server process you run yourself, so request data stays on your own infrastructure. The pages we read contain no statement about certifications, audits, data processing agreements or a security contact, and no company is named as an operator.
The configure page does cover handling of settings through environment variables and files, which is worth remembering when secrets sit in GUNICORN_CMD_ARGS. Security posture therefore depends on your own hardening, the reverse proxy in front of it and how quickly you apply new releases from the repository.
Gunicorn integrations and ecosystem
Integrations the vendor names (13):
- Django
- Flask
- FastAPI
- Pyramid
- Starlette
- Falcon
- Bottle
- Quart
- Nginx
- systemd
- Docker
- Gevent
- Eventlet
Fact sheet: Gunicorn
| Product | Gunicorn |
|---|---|
| Vendor website | gunicorn.org |
| Price visibility | Model recorded, no figure |
| Alternatives page | Not written yet |
| Headquarters | Not stated by the vendor |
| Primary guide | Best Application Servers in 2026 |
| Not in our data | funding rounds and staff numbers |
| Head-to-head pages | Not paired so far |
| Catalogue tags | WSGI, Pre-fork workers, MIT licence |
| Rank there | #8 of 14 |
Every guide that lists Gunicorn
| Guide | Position |
|---|---|
| Best Application Servers in 2026 | #8 of 14 |
What to look at instead of Gunicorn
These are the nearest neighbours of Gunicorn in the same list, picked on shared traits and price model.
- Apache TomEE (#5 of 14): Suits: Teams wanting CDI, EJB Lite and JPA on top of a familiar Tomcat base.
- Red Hat JBoss EAP (#12 of 14): Red Hat's supported Jakarta EE platform for on-premise and cloud. Watch out: the product page lists no Java or Jakarta EE versions, so the specification level has to be confirmed separately.
- Oracle WebLogic Server (#13 of 14): Oracle's Jakarta EE server in three editions, with 30-day cloud access. Watch out: the Standard Edition page names Jakarta EE 8 and 9.1, behind the Jakarta EE 11 targets of other servers in this category.
Gunicorn FAQ
9 answersWhat are the main alternatives to Gunicorn?
Start with Apache TomEE, Red Hat JBoss EAP and Oracle WebLogic Server. They are ranked and described in the same list.
Is Gunicorn free, or what does it charge?
As recorded: MIT-licensed package; the home page lists no prices or plans. On free access, no free tier or free plan is recorded.
Which country is Gunicorn from?
The vendor pages we reviewed state no country of establishment for Gunicorn, so we do not give one.
Who should choose Gunicorn?
Django, Flask or FastAPI teams on UNIX hosts deploying behind Nginx or systemd. Think twice if you are in this group: windows servers or teams wanting a commercial support contract.
How does Gunicorn rank for application servers?
Gunicorn is #8 of 14 in our application servers ranking.
Which Python version does Gunicorn need?
The install page says Python 3.12 or newer. Check your runtime before upgrading the package in an existing environment.
Does Gunicorn cost anything?
The home page lists no prices or plans, and the repository states the MIT License. Hosting and operation costs are your own.
Can it run async frameworks such as FastAPI?
FastAPI and Starlette appear in the framework list, and the home page names an ASGI worker using native asyncio among the worker types.
Who is behind the project?
The GitHub repository is owned by the user account benoitc. The pages we read name no company and give no country.
Gunicorn: sources
Read from the vendor’s own pages on 14 November 2026.
- gunicorn.org: Description, pre-fork model, worker types, framework list, documentation topics, no prices
- gunicorn.org/install: Python 3.12 requirement, install methods, Docker image, optional extras
- gunicorn.org/configure: Configuration sources, precedence, check and print commands, example config
- github.com/benoitc/gunicorn: MIT License, personal account ownership
How we rank
We test for four things rather than for features: setup time, price in practice, exit and audience. That puts Gunicorn at #8 of 14. The editorial process sets out how facts are gathered and how commercial links are handled.