Notably, Code Llama – Python 7B has outperformed Llama 2 70B on HumanEval and MBPP. All models have outperformed every other publicly available model on MultiPL-E, a testament to their superior ...