In this paper, we report a large-scale end-to-end language-independent multilingual model for joint automatic speech recognition (ASR) and language identification (LID). This model adopts hybrid CTC/attention architecture and achieves word error rate (WER) of 52.8 and LID accuracy of 93.5 on 42 languages with around 5000 hours of training data. We also compare the effects of using subword-level or character-level vocabulary for large-scale multilingual tasks. Furthermore, we transfer the pre-trained model to 14 low-resource languages. Results show that the pre-trained model achieves significantly better results than non-pretrained baselines on both language-specific and multilingual low-resource ASR tasks in terms of WER, which is reduced by 28.1% and 11.4% respectively.