Code Llama is an advanced large language model (LLM) tailored for code generation and discussing code, building on Llama 2. It comes in three versions: the foundational Code Llama, Code Llama – Python for Python coding, and Code Llama – Instruct for natural language instructions. Suitable for coding tasks like completion and debugging, it supports languages including Python, C++, Java, PHP, Typescript, C#, and Bash. Available in 7B, 13B, and 34B sizes, these models have been extensively trained on code data. The 7B and 13B models excel in code completion, while the 34B model offers superior coding assistance but may be slower.
