ANS is a new family of large alphabet accurate entropy coders (compression ratio like arithmetic coding), which offer many times better speeds than range coding - even faster than Huffman (PCS paper)
Its most interesting variants:
- tANS (tabled) - decoding step is very similar to Huffman (no multiplication), but it can operate on fractional bits (no longer limited to probabilities being a power of 1/2), e.g.
- tANS (range) - analogous to range coding, but using single multiplication per symbol instead of two, has single number state instead of two (perfect for SIMD), has more convenient frequency handling (perfect for adaptive coding), e.g.:

Used for example in
- Apple LZFSE = Lempel-Ziv + Finite State Entropy (tANS) for iOS9 - 3x faster, better compression than zlib,
- CRAM 3.0 of European Bioinformatics Institute: CRAM Benchmarks
and many others like Oodle LZNA (rANS), lzturbo (tANS), zhuff (tANS), ZSTD (tANS), LZA (rANS):

Here is an example of comparison for enwiki8 from :

All these compressors can be downloaded, ZSTD is open source ( ).