Meta LLaMA

open-source-llm-projects | foundation-models

Overview

Meta LLaMA is Meta's open-source large language model series. Starting with LLaMA 1 (2023), the family has expanded through LLaMA 2, LLaMA 3, and LLaMA 3.1, progressively increasing context length, reasoning ability, and open-source friendliness.

Key Models

• LLaMA 1: 7B–65B, first open LLM comparable to GPT-3
• LLaMA 2: 7B–70B, open weights with commercial license
• LLaMA 3: 8B–70B, improved reasoning, trained on 15T tokens
• LLaMA 3.1: 8B–405B, largest open weights model

Relationship to Other Projects

• Foundation for llama.cpp, vLLM, SGLang, and most open LLM ecosystems
• Compared with Mistral and Qwen as open model families

References

• Website: https://llama.meta.com/
• HuggingFace: https://huggingface.co/meta-llama