Meta Introduces Code Llama 70B, its Largest and Best-Performing Code Generation AI Model Yet

Meta has introduced an update to its code generation AI model, Code Llama 70B, which is touted as "the largest and best-performing model" in its lineup. Launched in August, Code Llama tools are available for free, catering to both research and commercial use. The latest version, Code Llama 70B, is designed to handle more queries, allowing developers to provide it with additional prompts during programming, resulting in increased accuracy.

Code Llama 70B achieved a 53 percent accuracy score on the HumanEval benchmark, surpassing GPT-3.5's 48.1 percent and approaching the 67 percent reported for GPT-4 in an OpenAI paper.

Built on Llama 2, Code Llama assists developers in generating code snippets from prompts and debugging human-written code. Last fall, Meta launched two other Code Llama tools, Code Llama - Python and Code Llama - Instruct, which focused on specific coding languages.

Code Llama 70B is available on three versions of the code generator and remains free for research and commercial purposes. The large model underwent training on 1TB of code and code-related data and is hosted on the code repository Hugging Face, providing access to GPUs for AI model execution.

Meta emphasized that its larger models, including 34B and 70B, yield superior results and offer improved coding assistance. Over the past year, other AI developers, including Amazon and Microsoft with GitHub Copilot, have also introduced code generators.