"In our own benchmark testing, Code Llama outperformed state-of-the-art publicly...

"In our own benchmark testing, Code Llama outperformed state-of-the-art publicly available LLMs on code tasks"