Rules of the participation:

• You can use any hardware, and there is no limit on inference time.
• You can use any open-weight model to provide its generations as submission.
• You can train your model on any data, including the data generated by commercial systems.
• After the competition ends you are required to publish used data and weights of the resulting model.
• We do not restrict the participation of the commercial systems, but they cannot receive the prize.

Please be aware of that some commercial machine translation or LLM services (namely, DeepL, Google Translate, ChatGPT, Claude, and Gemini) disallow the use of
their output for training other translation or AI models.

We are glad to share that Selectel became our sponsor! They provided us with 30000 bonuses for the winners of each competition in our track. This is enough to run A100 for two weeks in Selectel cloud.

You can join the competition now.

You can now join the competition.

We added more available datasets and models.