3 ms·

Chinese models are also awful with translations. Even the Deepseek R1 model performs worse than Mistral small.