Meta premier open-source code LLM family (7B to 70B parameters) specialized in synthesis, infilling, and private offline deployment.
Model weights open for free download under Meta Llama 2 Community License for research and commercial use
Code Llama is Meta flagship family of open-source large language models specifically fine-tuned for code synthesis and debugging, built on top of Llama 2. Offering 7B, 13B, 34B, and 70B parameter models, it features specialized variants including Code Llama - Python and Code Llama - Instruct. Capable of handling context windows up to 100k tokens, its breakthrough capability lies in native Fill-in-the-Middle (FIM) infilling, making it an exceptional self-hosted foundation for engineering teams demanding 100% offline code privacy and high-speed local inference.