Open Access (OA) literature is digital, online, free of charge, and free of
most copyright and licensing restrictions. What makes it possible is the
internet and the consent of the author or copyright-holder.
OA is entirely compatible with peer review, and all the major OA initiatives
for scientific and scholarly literature insist on its importance. Just as
authors of journal articles donate their labor, so do most journal editors
and referees participating in peer review.
OA literature is not free to produce, even if it is less expensive to produce
than conventionally published literature. The question is not whether
scholarly literature can be made costless, but whether there are better
ways to pay the bills than by charging readers and creating access barriers.
Business models for paying the bills depend on how OA is delivered.

There are two primary vehicles for delivering OA to research articles:
OA journals and OA archives or repositories.

OA Journals:
OA journals perform peer review and then make the
approved contents freely available to the world. Their
expenses consist of peer review, manuscript preparation,
and server space.
OA journals pay their bills very much the way broadcast
television and radio stations do: those with an interest
in disseminating the content pay the production costs
upfront so that access can be free of charge for everyone
with the right equipment. Sometimes this means that
journals have a subsidy from the hosting university or
professional society. Sometimes it means that journals
charge a processing fee on accepted articles, to be paid
by the author or the author’s sponsor (employer, funding
agency).
OA journals that charge processing fees usually waive
them in cases of economic hardship.
OA journals with institutional subsidies tend to charge no
processing fees.
OA journals can get by on lower subsidies or fees if
they have income from other publications, advertising,
priced add-ons, or auxiliary services. Some institutions
and consortia arrange fee discounts. Some OA publishers
waive the fee for all researchers affiliated with institutions
that have purchased an annual membership. There’s a lot
of room for creativity in finding ways to pay the costs of
a peer-reviewed OA journal, and we’re far from having
exhausted our cleverness and imagination.

OA Archives or repositories

OA archives or repositories do not perform peer review, but simply make their contents freely available to the world. They may contain unrefereed preprints, refereed post prints, or both.

Archives may belong to institutions, such as universities and laboratories, or disciplines, such as physics and economics.

Authors may archive their preprints without anyone else’s permission, and a majority of journals already permit authors to archive their post prints. When archives comply with the metadata harvesting protocol of the Open Archives Initiative, then they are interoperable and users can find their contents without knowing which archives exist, where they are coated, or what they contain. There is now open- source software for building and maintaining OAI-compliant archives and worldwide momentum for using it. The costs of an archive are negligible: some server space and a fraction of the time of a technician.

Reference: http://www.earlham.edu/~peters/fos/brief.htm