"Preprint" can mean two different things. A preprint can describe the state (or version) of a manuscript during the typical publication process in a journal. In this case, preprint refers to the version of the manuscript that is submitted to the publisher (see "Journal" route in figure below). During the publication process, this preprint version changes - often significantly - until it is accepted and finally published. In this scenario, the preprint is of no interest after publication and only the people directly involved in the publication process will ever see it.
The second meaning of preprint basically differs in one important aspect: This preprint is widely shared with the public while being published on a dedicated preprint server (see "Preprint server" route in the figure below). Such a preprint is also available for peer review and authors are free to implement changes to address reviewer comments. In parallel, this preprint can undergo the established publication process and eventually be published in a journal. However, this is not necessary and authors can decide to publish their work only as a preprint without ever publishing it in a journal.
Pros
For a publication in a journal, time from first submission until publication can be significant; several months and even more than a year are no exception. In contrast, a preprint can be published within hours to days. This has important advantages for you as an author:
Preprints are increasingly accepted by funding agencies and hiring committees when assessing candidates. This can be particularly relevant for early career scientists. Instead of waiting many months until your work is published in a journal, you can share a preprint as direct proof of the state and quality of your work. Version control also allows you to present and document the current state of your work to the evaluating committee now and to update the manuscript with new results later on.
We are currently not aware of any established preprint server (see below) that requires authors to pay for publishing their preprint. Thus, preprints enable you to publish your work at very low costs.
On established preprint servers, copyright remains with the authors and all preprints are published Open Access. Typically, you can choose between different licensing options, in particular among the different CC licenses (see our explanation about CC licenses).
Preprints can be easily updated whenever authors choose to do so. Different versions of a manuscript keep the same DOI (see our explanation about DOI) and are displayed together (see for example "Info/History" of this article).
Journals often have very specific requirements for the formatting of their articles. These requirements can sometimes collide with the best way for presenting and explaining your data. Preprint servers, in contrast, have very few such requirements and allow you to format your manuscript according to your own needs.
Cons
Publishers can consider a preprint on a preprint server as "prior publication" and, therefore, decline to publish the same work in their journals. However, almost all publishers agreed by now that publishing a preprint on a preprint server does not interfere with publishing in their journals (see here for an overview of publisher policies). You can also contact the editorial office of your journal of choice to be completely sure about their preprint policy.
Formal peer review is not part of the preprint server publication process. However - while rare - preprints can also receive feedback from peers in various form and authors are free to react to this feedback and to change their manuscript.
It should also be noted that the quality of peer review in journals differs widely and that peer review by itself is no indication for the quality of a manuscript, be it for journal publications or for preprints.
While funders and hiring committees often consider preprints in their evaluation, it is probably safer to assume that peer reviewed publications in journals are still considered the gold standard publication format for scientific work. However, this can be highly field dependent and change rapidly.
Many journals allow citing preprints. However, there is no consensus on how these citations count if the cited preprint is eventually published in a journal.
Google Scholar lists preprints and allows to combine separate entries for a preprint and its journal-published version into one entry (including their citations). This is currently the exception among citation databases and there is a risk that citations of a preprint do not count for the journal-published version of the same content.
The number of preprint servers is increasing, the field is very dynamic and they rely on external funding in absence of publication fees. Thus, there is a risk that a preprint server goes out of business and that published preprints are lost. To avoid this risk, we recommend choosing among the established preprint servers (see below) for which sustainability is ensured.
Preprint servers
arXiv is one of the oldest and most established preprint servers with more than 2'500'000 published preprints. The thematic focus is on physics and mathematics but has expanded to quantitative topics in other sciences.
bioRxiv kickstarted preprints in the life sciences. It started only in 2013 but hosts nowadays more than 250'000 preprints. Compared to arXiv, bioRxiv is somehwhat restrictive regarding manuscript formats and does not accept review articles and other manuscripts without original data.
ChemRxiv is co-owned and collaboratively managed by several important chemical societies.
medRxiv is the "cousin" of bioRxiv and hosts preprints that are more focussed on medical topics, including medical trials.
TechRxiv launched in 2020 and is owned by the Institute of Electrical and Electronics Engineers (IEEE).
Zenodo, operated by CERN, hosts not only manuscripts but also presentations, data sets, research software, ... While being hosted by CERN, Zenodo has no thematical focus and accepts manuscripts irrespective of the topic and the authors' affiliations.
