Code Llama: serious open-source coding model
In one sentence Meta releases Code Llama (7B, 13B, 34B), a code-specialized fine-tune of Llama 2. Three variants per size: base, Python-specific, instruction-tuned. Llama 2 commercial license.
Meta ships Code Llama, a version of Llama 2 trained specifically to write and understand code. Three sizes (7, 13, 34 billion parameters) and three variants: base, Python-specific, and an "instruct" one that understands natural-language commands.
It's the first serious open-source alternative to Codex / GPT-4 for coding: runs locally (even on a good consumer GPU), supports Copilot-style autocomplete, ships with a permissive commercial license.
For anyone wanting a self-hosted Copilot, there's finally a solid foundation.
Companies
Meta
Tools
Code Llama
Tags
Sources