I've released a new markdown library on Hackage: http://hackage.haskell.org/package/cheapskate This library is designed to be used in web applications. It is small, accurate, and fast, in pure Haskell with few dependencies. All output is sanitized through a whitelist by default. It is designed to have performance that varies linearly with the input size, even with garbage input. To illustrate: % head -c 100000 /dev/urandom | iconv -f latin1 -t utf-8 | time cheapskate >/dev/null cheapskate > /dev/null 0.15s user 0.01s system 82% cpu 0.199 total % head -c 1000000 /dev/urandom | iconv -f latin1 -t utf-8 | time cheapskate >/dev/null cheapskate > /dev/null 1.53s user 0.03s system 88% cpu 1.770 total % head -c 10000000 /dev/urandom | iconv -f latin1 -t utf-8 | time cheapskate >/dev/null cheapskate > /dev/null 15.50s user 0.20s system 89% cpu 17.632 total This is a test that many markdown parsers fail (including my own pandoc and the markdown package on Hackage). Performance is about seven times faster than pandoc (with five times less memory used), and about 25% faster than the markdown package on Hackage. Generic functions are provided that allow transformations of the AST prior to rendering (e.g., promotion of headers, insertion of syntax highlighting, or the conversion of specially marked code blocks into diagrams). Several markdown extensions are supported, and sensible decisions have been made about several aspects of markdown syntax that are left vague by John Gruber's specification. For details, see the README at https://github.com/jgm/cheapskate.