The Tatoeba MT Challenge Models - newstest2021 benchmarks (source language plot)

• Select a target language from the menu on the top right corner of the map
• The map shows source languages that are supported by available models
• Colors indicate the quality on a scale from red (bad) to green (best) measured by chr-F2 scores
• The size of the dot represents the size (i.e. reliability) of the test set
• Click on the dots to get more information about the language pair and to get links for downloading the model
• Maps for other benchmarks
• For the opposite translation direction: Click on the target language plot