In model railroading, the term freelance is used to mean something that is not a model of a specific prototype, but rather something that never existed in reality - something fictional. The term is generally not used when referring to something that attempts to be a model of some real-world prototype - regardless of the accuracy or the compromises that had to be made. Antagonism between prototype modellers and freelancers is a long-standing part of the hobby.
There are many types of freelancing and many reasons for doing so. One important reason is that by freelancing, one avoids compromise. Railroads are fundamentally hard to model - it's an attempt to fit a huge, linearly based thing into a much smaller, even in scale distances, area, and one quite differently shaped. Prototype trains are often way too long to accurately model. Real railroads have hundreds, even thousands, of locomotives, and maybe tens of thousands of cars. Accurate prototype information, especially for earlier historical periods, is hard to obtain, incomplete, and conflicting.
By freelancing, the model railroader avoids having to compromise nearly as much, by the simple fact that 'the way it is' can be laid down by modeler's decree. The shape of one's modelling space becomes the shape of the landscape of one's invented railroad. If only one locomotive is all that the money can stretch to, well, then this railroad only had one locomotive. Want to run 1880s locomotives next to 1950s vintage? Then that's the way it is on this railroad. The modeler is freed from the work of having to find a prototype that matches up to their restraints, or the frustrating challenge of having to compromise to fit within space and budget.
This becomes especially tempting when modeling what interests you would just be too hard - when accurate models of specific prototypes are hard to get. An example would be modeling nineteenth century railroads, for which few models are available.
All freelancing, of course, has to be an acceptable suspension of disbelief for the individual sculpting it. There's always an implied 'Could it have happened this way?' test, in the modeler's mind. If one didn't know better, would one accept that this fictional railroad really exists, having just lurked out of sight and mind? That's why freelancing is more popular in the United States, with its myriad railroads, than in Europe, where most countries have national rail systems and it's not credible for there to be another railroad. The possibility, though, of there being a narrow gauge line in Wales that one never heard of, or a secondary carrier in the American Midwest, or a streetcar line in some unheard of city in eastern Europe - those don't fail the credibility test.
Freelance model railroads can be just as serious as one fully modeled on a real prototype - and can be just as hard, if not harder, work. Some modelers go to the extent of creating whole books of rules and regulations for their railroads, operating practices, timetables, and the like -- things the prototype modeler can just find pre-made with a little research. These modelers tend to be ones more interested in railroad practice and operation than the specific details of a real-life railroad - the challenge of planning their own, realistic as possible, railroad empire is what appeals. Such people sometimes call their creations protofreelance to emphasize that while the railroad itself is fictional, it's firmly grounded in real prototype practice.
Some modelers choose to freelance elements of their model railroad, but not all of it. One common way to do this is to model a real railroad, with its real locomotives and cars, but to invent a fictional location - either completely out of thin air, or modelling a proposed but never built line, or one that in real life was closed down before the era in question. This allows the freelancer's latitude of shaping the model railroad within the confines of a space, while still modeling a railroad and equipment that interests one.