Installing *BSD in 2025 part 0a - Modern myths: AI ramblings (DragonFly BSD and FreeBSD)
In this series I plan to take a look at the installers of the various BSD operating systems. There's a lot of good reasons to install a BSD on your computer(s); and while that does not depend on the particular year too much, it's still interesting to take a look at the state of this OS family now and then.
We all know the popular "BSD is dying" myth that has been with us for many years (or even decades) now. Not too surprisingly, there are many other myths as well as misconceptions floating around. Since it's 2025 and people are using Large Language Models ("AIs") more and more, I was curious how well informed about *BSD they are. This is the first half of part 0 which covers DragonFly BSD and FreeBSD.
Please keep in mind what so-called "AI" assistants do: They allow you to _simulate_ a chat where you get responses to your prompts based on what - according to the training data that was used to create the model - _a human would likely have replied_. So while these responses are usually pretty good in terms of correct language (which makes them look authentic), they can actually be utterly nonsensical at the same time.
When the training of a model has not covered a particular topic, that model will make things up ("hallucinate"). If a model's answers are mostly correct and precise we can assume that the training data contained a fair share of data on this matter. If it makes mostly false claims the opposite is probably true.
This is by no means scientific research on BSD coverage in common AI models, but I decided to ask a couple of LLMs two questions (I initially had more but this is too much material) for each one of the four major BSDs:
I'm not deliberately trying to get the most weird / funny results. Therefore when possible, I picked the larger versions of models that have more parameters and thus potentially more "knowledge". I also tried to include some from high parameter models to smaller ones. These are the models that I tested:
For convenience I used a more powerful Linux machine that I have access to for the bigger three but the other half were run on one of my FreeBSD laptops (though admittedly using a Linuxulator jail to get more recent versions of ollama than what we have in ports).
The first thing that I asked the chat bots about was to summarize each of the BSD operating systems. Let's see what they came up with!
Llama 3.3 summarizes the project like this:
"The DragonFly BSD project is a free and open-source operating system that forked from FreeBSD 4.8 in 2003, with a focus on providing a highly scalable, secure, and performant platform, particularly in areas such as clustering, high-performance networking, and storage, through the use of advanced technologies like kernel-level threading and a lightweight, message-passing architecture. DragonFly BSD is designed to be a modern, agile operating system that can take advantage of today's multi-core processors and large amounts of memory, while still maintaining the stability and simplicity that are hallmarks of the BSD family, although its smaller user base and more experimental nature set it apart from other BSDs."
Llama 3.3's response really shines: It contains quite a bit of relevant information and even got the numbers right (both the version that dfly forked from and the year)! I'm not sure where the "agile OS" part comes from, but especially pointing out that it has a rather small user base is both accurate and important to know.
Mixtral's summary goes like this:
"DragonFly BSD is an open-source, innovative, and UNIX-like operating system built on a single-kernel design with a focus on performance, scalability, and SMP (Symmetric Multi-Processing) capabilities. Its unique features include the Lightweight Kernel Threads (LWKT), an advanced scheduling mechanism that promotes fairness and performance across multi-core systems. DragonFly BSD is suitable for high-performance computing, embedded devices, and educational purposes, particularly appealing to users who want to explore modern kernel design principles while benefiting from a robust and user-friendly system."
Pointing out today that the OS is SMP capable makes little sense, since that basically goes without saying. It is however true that dfly was created due to its founder, Mathew Dillon, and most other FreeBSD developers disagreeing over the SMP approach of the kernel. So while it's not important information, it's not wrong, either. Mentioning LWKT on the other hand, is great since that's one of dfly's characteristics. The claim that it is suitable for embedded devices is something that I would question, though, since DragonFly BSD supports only the amd64 platform (and even there focuses on specific processor families) with no support for ARM, RISC-V or anything. I'm not going to judge its user-friendliness since that's a matter of taste, but I will say that to me it doesn't look like that's a particularly important goal of the project.
According to Exaone,
"DragonFly BSD is a Unix-like operating system that emphasizes scalability, performance, and a unique microkernel architecture inspired by Mach. It focuses on modernizing BSD with a distributed locking manager and a HAMMER file system designed for reliability and speed. Targeted at users who value cutting-edge features and a robust development model, DragonFly BSD offers a distinct alternative to traditional BSD systems, appealing to seasoned users looking for innovative approaches to system architecture and performance optimization."
In Exaone's answer we have the first false (or at least very imprecise) claim: DragonFly BSD is _not_ using a microkernel design. It actually uses a hybrid kernel which is partially inspired by microkernel concepts (I have no idea about how much Mach has or has not been particularly influential, though). I don't think I would have described these two characteristics of HAMMER as the most important ones, but it's not exactly wrong.
Gemma 2 described the OS like this:
"DragonFlyBSD is a Unix-like operating system forked from FreeBSD that aims to innovate and improve upon the traditional BSD design. It focuses on delivering high performance and scalability through a redesigned kernel architecture, making it well-suited for demanding workloads and modern hardware."
Short and sweet but not very informative.
Phi 4 answered the prompt like this:
"DragonFly BSD is an operating system that evolved from FreeBSD, designed with a focus on scalability and performance in multi-core processors. It introduces several architectural innovations, including per-CPU data structures, which enhance parallel processing capabilities and reduce contention among cores. The project emphasizes modularity, allowing for flexible kernel configurations, and has its unique take on handling tasks such as scheduling, networking, and file systems with the HAMMER2 filesystem being a notable feature. DragonFly BSD also aims to improve upon existing technologies by refining userland tools and utilities, striving for simplicity in design while maintaining compatibility with FreeBSD's legacy where beneficial. The system is appreciated for its stability and efficiency, particularly in server environments that demand high-performance computing and fast network throughput."
I will admit that I don't know about the "per-CPU data structures" and think that the "maintaining compatibility with FreeBSD's legacy where beneficial" is a bit of a filler. Otherwise it sounds about right!
Here is Glm 4's take:
"DragonFly BSD is an open-source operating system that diverged from the FreeBSD project in 2003, focusing on improved concurrency and cutting-edge features like multithreading and asynchronous I/O. It's valued by experienced Unix users for its commitment to technical excellence, advanced file system architecture, and innovative design choices, offering a modern, reliable alternative with roots in the BSD heritage."
While I wouldn't call multithreading a "cutting-edge feature", it's true that dfly did quite a bit of work in these areas. The description is not half bad!
As expected, the biggest model was able to deliver a response with a lot of facts, however the smallest model surprised me by also providing a quite nice and informative description. Some of the models in between also provide good answers but some of them contain false claims. Gemma 2 is notable for delivering a very short description - but it at least managed to not make any mistakes. I was also surprised about some of the models obviously having more than just very vague information about this operating system which is pretty nice!
Here's Llama 3.3's answer:
"The FreeBSD project is a free and open-source operating system based on the BSD codebase, with a focus on performance, stability, and scalability, making it a popular choice for servers, embedded systems, and desktops, particularly in environments that require high networking throughput and reliability. Unlike other BSDs, FreeBSD has a more permissive licensing model and a larger user base, which contributes to its popularity and widespread adoption in industries such as networking, storage, and virtualization, with many companies using it as the foundation for their products, including Apple's macOS and iOS operating systems."
While the machine that I write this post on runs FreeBSD, I wouldn't say that the FreeBSD desktop is _popular_. The FreeBSD foundation has been doing a fair bit of work towards making FreeBSD a great experience on the desktop, but we definitely haven't reached the point where the OS would compare favorably to Linux in most regards. It does in several and that's why I clearly prefer it, but there's valid reasons for why desktop FreeBSD is more of a niche thing. The license claim is misleading, too: All of the BSDs are of course BSD-licensed. The claim is probably based on the fact that there are several BSD license (4-clause, 3-clause and 2-clause BSD) and FreeBSD pioneered the most permissive of them. Apple indeed uses several tools that are derived from the FreeBSD userland, but "macOS is based on FreeBSD" is a popular myth (it uses an entirely different kernel and kernel architecture for example).
Mixtral sums up the OS as follows:
"FreeBSD is an open-source, general-purpose UNIX-like operating system known for its stability, performance, and extensive integration of userland applications. It combines a modern and feature-rich kernel with a well-organized ports collection, offering thousands of ready-to-build software packages. FreeBSD caters to servers, desktops, and appliances alike, making it a popular choice for enterprises, educational institutions, and enthusiasts who appreciate its balance between innovation and reliability."
Compared to the more minimalistic approach of the other BSDs, one could probably FreeBSD's userland application integration as extensive - bit it's still a fairly lean system. The other information is correct, too.
Exaone's take sounds like this:
"FreeBSD is a powerful, open-source Unix-like operating system known for its scalability, performance, and robustness, widely used in server environments and network infrastructure. It offers a stable and efficient platform with advanced features like ZFS for storage management and a well-optimized kernel, making it a preferred choice for users who require high performance and reliability, similar to what seasoned Linux users appreciate but with unique strengths in networking and modern file system capabilities."
This one is perfectly fine. While I would prefer to describe it differently I don't disagree with the answer.
Here's what Gemma 2 says:
"FreeBSD is a powerful and versatile Unix-like operating system known for its stability, performance, and strong community support. Similar to Linux distributions but built on the BSD codebase, it excels in server environments and provides a rich set of features for both general and specialized use cases."
This, too, is a pretty short description, but it's correct.
Phi 4 describes the system like this:
"FreeBSD is an advanced open-source Unix-like operating system known for its performance, scalability, and advanced networking capabilities. It provides a rich set of features and tools for managing complex network infrastructures, making it popular in enterprise environments for servers and storage solutions. The project emphasizes stability and high-quality code, with an active community contributing to its extensive ports collection that allows easy installation of software packages. FreeBSD is also recognized for its pioneering work in Jails, a technology providing lightweight process isolation similar to containers, which enhances system security and efficiency. Additionally, it includes advanced features like ZFS filesystem support, integrated DHCP server, and robust firewall capabilities, making it versatile for a wide range of applications from desktops to high-performance computing environments."
Overall, Phi 4 delivers a pretty good answer, pointing to some of the notable characteristics of the system. However there's one mistake in there: While FreeBSD of course comes with a DHCP client, a server is not part of the base system and would need to be installed on top. While at some point the system came with a DNS server (BIND), I don't think a DHCP server was ever build-in in any previous version.
Glm 4's answer goes like this:
"FreeBSD is an advanced Unix-like operating system derived from BSD, known for its stability, performance, and robust networking capabilities. It appeals to seasoned Linux users with its strong security model, flexible licensing, extensive hardware support, and rich set of features, offering a familiar, yet distinct, alternative to Linux environments."
While that's not too precise, I'd at least say it's correct and probably sufficient as a quick summary.
At first glance, it is easier to describe FreeBSD than DragonFly BSD since it's much more popular and there's a lot more information on it available online. However this also increases the probability of confusion, and exactly that's what I found has happened. In general I'd say that most of the replies were okay or even good, though.
The second task was to tell me more about the distinctive features of the respective system. I will not put all the output in here as that would be too much for a blog post. Instead I'm going to concentrate on the list of points they provided and comment on that. If the point was accurate and I don't have any comments, I just present the bullet point.
Llama 3.3 lists 10 features:
Among the suggested use cases are "Cloud infrastructure and virtualization" as well as "Embedded systems" which I'd say are not exactly what dfly is being used for. I also don't think it's being used in HPC (High-performance computing) clusters (would be interesting, though!).
The conclusion is correct, though: "It's worth noting that DragonFly BSD is still a relatively niche operating system compared to other BSDs and Linux distributions, but it has a dedicated community and a strong focus on innovation and performance."
Mixtral points out 9 features of the OS:
Exaone lists 9 features:
Short and rather general like before, Gemma 2 only offers these 4 points (all of which are explained in one or two sentences which don't offer much additional information):
Phi 4 offers 8 features to educate us about DragonFly BSD:
Glm 4 describes the OS with 10 points:
So when it comes to the knowledge of features present in DragonFly BSD, we're seeing a mixed bag. There's some correct points, some that are debatable or inaccurate, but also quite a few downright false claims. Since dfly is the least popular of the 4 main BSD operating systems this was to be expected, though.
Llama 3.3 sums up FreeBSD's features in 10 points:
Mixtral lists 8 points when asked for FreeBSD's features:
Exaone offers 9 points for FreeBSD's features:
Once again, Gemma 2 keeps it brief with 4 list items:
Phi 4's take contains 8 points:
Finally Glm 4 provides 10 points in its response:
When it comes to FreeBSD, the results were pretty much ok in general with a few exceptions in both directions. Some LLMs provided information that's actually good while others included claims that are completely wrong.
_Always double-check answers that LLMs generated_ is the obvious one. This is not limited to *BSD - well, and it can apply to real people as well! The more niche a topic is, the less likely it becomes that a lot of material on it was selected for training data used to create LLMs. The assumption that FreeBSD as the most popular OS of the family is also more widely covered than the least popular one, DragonFly BSD, is confirmed. But while there's more weird claims about the latter OS the first is not save from this phenomenon, either.
I've asked all the models the same questions about Linux as well - and in fact got _zero_ wrong answers! Ok, Mixtral made one claim regarding the licensing that is imprecise ("Linux is distributed under a free and open-source license, which means that anyone can view, modify, and distribute its source code without any legal or financial restrictions") since the GPL for example restricts your right to keep the source to yourself if you distribute the compiled object. But that's really it: All other statements were correct.
In the second part of this post, we'll take a look at what the same LLMs may tell us about the other two major BSDs: NetBSD and OpenBSD.
Installing *BSD in 2025 part 0a – Modern myths: AI ramblings (NetBSD and OpenBSD)