Meta Open Sources Code Llama to Compete in AI Coding Space
Meta releases Code Llama, an open source AI model designed to generate, explain, and debug code across popular programming languages. The new system builds directly on the company's recently released Llama 2 foundation.
Meta releases Code Llama, an open source artificial intelligence model that generates and explains computer code in natural language. This new system directly competes with established tools like GitHub Copilot and Amazon CodeWhisperer by completing and debugging software across multiple popular programming languages, including Python, C++, Java, and PHP.
The model builds directly on Meta's previously released Llama 2 foundation but focuses heavily on code-specific training. While Llama 2 possesses some basic coding abilities, Code Llama receives specialized training on 500 billion tokens of code and code-related data to significantly improve its programming quality and accuracy.
Meta offers Code Llama in several distinct variations to serve different developer needs, including a version specifically optimized for Python and another fine-tuned to understand natural language instructions. By open sourcing this technology, Meta aims to encourage community evaluation, rapid innovation, and the identification of security vulnerabilities within the AI models.