diff --git a/benchmarks/benchmark_batches.pl b/benchmarks/benchmark_batches.pl
new file mode 100644
index 0000000..ba6c50c
--- /dev/null
+++ b/benchmarks/benchmark_batches.pl
@@ -0,0 +1,245 @@
+#!/usr/bin/env perl
+use strict;
+use warnings;
+use Benchmark qw!:hireswallclock :all!;
+use Data::Dumper;
+use POSIX 'round';
+
+our @ARGV;
+
+# The first parameter is the number of iterations.
+
+my $FILE = 'effi-1x-utf8.txt';
+system 'gzip -dkf ./corpus/' . $FILE . '.gz';
+my $iter = 1;
+
+if ($ARGV[0]) {
+  $iter = $ARGV[0] + 0;
+};
+
+my $temp_dir = '/euralex/batches/';
+mkdir($temp_dir);
+
+# Read file
+my $single = '';
+if (open(my $f, '<', './corpus/' . $FILE)) {
+  while(!eof($f)) {
+    $single .= <$f>;
+  };
+  close($f);
+};
+
+warn "Concatenate file\n";
+
+my $count = 300;
+
+# Concat $count times
+my $data = '';
+foreach (1..$count) {
+  $data .= $single;
+};
+
+# save file in docker
+open(X, '>', $temp_dir . 'effi-' . $count . 'x-utf8.txt') or die $!;
+print X $data;
+close(X);
+
+warn "Calculate offsets\n";
+
+# Calculate offsets with datok (just because) for segmenting
+my $offsets = `cat ${temp_dir}effi-${count}x-utf8.txt | ./Datok/datok tokenize -t ./Datok/testdata/tokenizer.matok --no-tokens --no-sentences -p -`;
+my @offsets = split / /, $offsets;
+
+unlink './corpus/effi-' . $count . 'x-utf8.txt';
+
+$FILE = undef;
+
+my $models = {
+  'wc' => sub {
+    system 'wc -w ./corpus/'.$FILE.' > /dev/null';
+  },
+  'SoMaJo' => sub {
+    system 'somajo-tokenizer '.$temp_dir.$FILE.' --split_sentences > /dev/null';
+  },
+  'SoMaJo_p2' => sub {
+    system 'somajo-tokenizer '.$temp_dir.$FILE.' --parallel=2 --split_sentences > /dev/null';
+  },
+  'SoMaJo_p4' => sub {
+    system 'somajo-tokenizer '.$temp_dir.$FILE.' --parallel=4 --split_sentences > /dev/null';
+  },
+  'SoMaJo_p8' => sub {
+    system 'somajo-tokenizer '.$temp_dir.$FILE.' --parallel=8 --split_sentences > /dev/null';
+  },
+  'Datok_matok' => sub {
+    system 'cat '.$temp_dir.$FILE.' | ./Datok/datok tokenize -t ./Datok/testdata/tokenizer.matok - > /dev/null'
+  },
+  'Datok_datok' => sub {
+    system 'cat '.$temp_dir.$FILE.' | ./Datok/datok tokenize -t ./Datok/testdata/tokenizer.datok - > /dev/null'
+  },
+  'OpenNLP_Simple' => sub {
+    system 'cat '.$temp_dir.$FILE.' | ./opennlp/bin/opennlp SimpleTokenizer > /dev/null';
+  },
+  'OpenNLP_Tokenizer_de-ud-gsd' => sub {
+    system 'cat '.$temp_dir.$FILE.' | ./opennlp/bin/opennlp TokenizerME ./opennlp/models/opennlp-de-ud-gsd-tokens-1.0-1.9.3.bin > /dev/null';
+  },
+  'OpenNLP_Sentence_de-ud-gsd' => sub {
+    system 'cat '.$temp_dir.$FILE.' | ./opennlp/bin/opennlp SentenceDetector ./opennlp/models/opennlp-de-ud-gsd-sentence-1.0-1.9.3.bin > /dev/null';
+  },
+  'TreeTagger' => sub {
+    system 'cat '.$temp_dir.$FILE.' | perl ./treetagger/cmd/utf8-tokenize.perl -a ./treetagger/lib/german-abbreviations > /dev/null';
+  },
+  'deep-eos_bi-lstm-de' => sub {
+    system 'python3 ./deep-eos/main.py --input-file '.$temp_dir.$FILE.' --model-filename ./deep-eos/bi-lstm-de.model --vocab-filename ./deep-eos/bi-lstm-de.vocab --eos-marker "§" tag > /dev/null';
+  },
+  'deep-eos_cnn-de' => sub {
+    system 'python3 ./deep-eos/main.py --input-file '.$temp_dir.$FILE.' --model-filename ./deep-eos/cnn-de.model --vocab-filename ./deep-eos/cnn-de.vocab --eos-marker "§" tag > /dev/null';
+  },
+  'deep-eos_lstm-de' => sub {
+    system 'python3 ./deep-eos/main.py --input-file '.$temp_dir.$FILE.' --model-filename ./deep-eos/lstm-de.model --vocab-filename ./deep-eos/lstm-de.vocab --eos-marker "§" tag > /dev/null';
+  },
+  'JTok' => sub {
+    chdir '/euralex/JTok/bin';
+    system 'sh tokenize '.$temp_dir.$FILE.' de > /dev/null';
+    chdir '/euralex';
+  },
+  'KorAP-Tokenizer' => sub {
+    system 'cat '.$temp_dir.$FILE.' | java -jar ./KorAP-Tokenizer/KorAP-Tokenizer.jar -l de -s > /dev/null'
+  },
+  Syntok_tokenizer => sub {
+    system 'python3 -m syntok.tokenizer '.$temp_dir.$FILE.' > /dev/null';
+  },
+  Syntok_segmenter => sub {
+    system 'python3 -m syntok.segmenter '.$temp_dir.$FILE.' > /dev/null';
+  },
+  Waste => sub {
+    system 'cat '.$temp_dir.$FILE.' | waste -N -v0 --rcfile=./Waste/waste.rc > /dev/null';
+  },
+  nnsplit => sub {
+    system './nnsplit/nnsplit_bench '.$temp_dir.$FILE.' > /dev/null'
+  },
+  elephant => sub {
+    system './elephant-wrapper/bin/tokenize.sh -i '.$temp_dir.$FILE.' UD_German > /dev/null'
+  },
+  cutter => sub {
+    system 'python3 ./cutter/cutter.py nosent '.$temp_dir.$FILE.' > /dev/null'
+  },
+  blingfire_tok => sub {
+    system 'python3 ./blingfire/blingfire_tok.py '.$temp_dir.$FILE.' > /dev/null'
+  },
+  blingfire_sent => sub {
+    system 'python3 ./blingfire/blingfire_sent.py '.$temp_dir.$FILE.' > /dev/null'
+  },
+  spacy_tok => sub {
+    system 'python3 ./spacy/spacy_tok.py '.$temp_dir.$FILE.' > /dev/null'
+  },
+  spacy_dep => sub {
+    system 'python3 ./spacy/spacy_sent.py dep '.$temp_dir.$FILE.' > /dev/null'
+  },
+  spacy_stat => sub {
+    system 'python3 ./spacy/spacy_sent.py stat '.$temp_dir.$FILE.' > /dev/null'
+  },
+  spacy_sentencizer => sub {
+    system 'python3 ./spacy/spacy_sent.py sentencizer '.$temp_dir.$FILE.' > /dev/null'
+  },
+  Stanford => sub {
+    system 'CLASSPATH=/euralex/stanford-corenlp-4.4.0/* java edu.stanford.nlp.pipeline.StanfordCoreNLP ' .
+      '-props german -annotators tokenize,ssplit,mwt -tokenize.language=german -file '.$temp_dir . $FILE
+  },
+  Stanford_t2 => sub {
+    system 'CLASSPATH=/euralex/stanford-corenlp-4.4.0/* java edu.stanford.nlp.pipeline.StanfordCoreNLP ' .
+      '-props german -annotators tokenize,ssplit,mwt -tokenize.language=german -threads=2 -file '.$temp_dir . $FILE
+    },
+  Stanford_t4 => sub {
+    system 'CLASSPATH=/euralex/stanford-corenlp-4.4.0/* java edu.stanford.nlp.pipeline.StanfordCoreNLP ' .
+      '-props german -annotators tokenize,ssplit,mwt -tokenize.language=german -threads=4 -file '.$temp_dir . $FILE
+    },
+  Stanford_t8 => sub {
+    system 'CLASSPATH=/euralex/stanford-corenlp-4.4.0/* java edu.stanford.nlp.pipeline.StanfordCoreNLP ' .
+      '-props german -annotators tokenize,ssplit,mwt -tokenize.language=german -threads=8 -file '.$temp_dir . $FILE
+    },
+  Stanford_tokonly => sub {
+    system 'CLASSPATH=/euralex/stanford-corenlp-4.4.0/* java edu.stanford.nlp.pipeline.StanfordCoreNLP ' .
+      '-props german -annotators tokenize -tokenize.language=german -file '.$temp_dir . $FILE;
+    unlink $temp_dir . $FILE . '.out';
+  },
+};
+
+#delete $models->{'wc'};
+#delete $models->{'SoMaJo'};
+#delete $models->{'SoMaJo_p2'};
+#delete $models->{'SoMaJo_p4'};
+#delete $models->{'SoMaJo_p8'};
+#delete $models->{'Datok_matok'};
+#delete $models->{'Datok_datok'};
+#delete $models->{'OpenNLP_Simple'};
+#delete $models->{'OpenNLP_Tokenizer_de-ud-gsd'};
+#delete $models->{'OpenNLP_Sentence_de-ud-gsd'};
+#delete $models->{'TreeTagger'};
+#delete $models->{'deep-eos_bi-lstm-de'};
+#delete $models->{'deep-eos_cnn-de'};
+#delete $models->{'deep-eos_lstm-de'};
+#delete $models->{'JTok'};
+#delete $models->{'KorAP-Tokenizer'};
+#delete $models->{'Syntok_tokenizer'};
+#delete $models->{'Syntok_segmenter'};
+#delete $models->{'Waste'};
+#delete $models->{'nnsplit'};
+#delete $models->{'elephant'};
+#delete $models->{'Stanford'};
+#delete $models->{'Stanford_t2'};
+#delete $models->{'Stanford_t4'};
+#delete $models->{'Stanford_t8'};
+#delete $models->{'Stanford_tokonly'};
+#delete $models->{'cutter'};
+#delete $models->{'spacy_tok'};
+#delete $models->{'spacy_sentencizer'};
+#delete $models->{'spacy_dep'};
+#delete $models->{'spacy_stat'};
+#delete $models->{'blingfire_tok'};
+#delete $models->{'blingfire_sent'};
+
+
+my $t0 = Benchmark->new;
+
+# Get some batch files with
+# 1000 tokens
+# 2000 tokens
+# 4000 tokens
+# 8000 tokens
+# ...
+# 8192000 tokens
+for (my $x = 0 ;$x < 100;$x++) {
+  my $i = (2**$x) * 1000;
+
+  warn "Create $i batch\n";
+
+  last if ($i*2)+1 > scalar(@offsets);
+
+  my $off = $offsets[($i*2)+1];
+  last unless $off;
+
+  $FILE = 'effi-batch-' . $i . '.txt';
+
+  open(Y, '>', $temp_dir . $FILE);
+  print Y substr($data, 0, $off);
+  close(Y);
+
+  my $cmp = timethese($iter => $models);
+
+  print "\n----------------------------------\n";
+
+  foreach my $tool (sort keys %$cmp) {
+    my $seconds_per_run = $cmp->{$tool}->[0] / $cmp->{$tool}->[5];
+    my $tokens_per_msecond = ($i / $seconds_per_run) / 1000;
+    print $tool,"\t",$i,"\t", $seconds_per_run, "\t", sprintf("%.2f", $seconds_per_run), "\t", $tokens_per_msecond, "\t", sprintf("%.2f", $tokens_per_msecond), "\n";
+  };
+
+  print "\n----------------------------------\n";
+
+  unlink $temp_dir . $FILE;
+};
+
+
+
+print "Benchmarking took: ", timestr(timediff(Benchmark->new, $t0)), "\n";
+