Open weights from Meta
Llama is the series of large language models developed and published by Meta. What matters most is that the weights are published and, subject to the license, free for commercial use. As the leading open LLM family it has become the foundation for research and development worldwide.
Why it mattered
After ChatGPT, it looked as though capable LLMs would remain the private assets of a few companies. The 2023 release of Llama changed that. Once near-frontier models could be run and modified by anyone, local LLMs, in-house corporate AI, research, and derivative models all opened up at once.
- Download and run it on your own server or PC, keeping data in-house
- Customize freely through fine-tuning or LoRA
- Countless derivatives, forming the center of the open ecosystem
What “open” means here
Llama is released under its own license with conditions including usage scale. Whether to call it open source is debated, and open weights is the more accurate description. In practice, it remains freely usable by most companies and individuals.
Context worth having
Each new Llama release is read against how far the gap to closed models — GPT, Claude, Gemini — has narrowed. Competition from other open families, such as Mistral in France and several from Chinese labs, is also active. As the standard-bearer for open AI, it belongs alongside Hugging Face and local LLMs.
The main reason organizations choose Llama is using a capable model without sending their data outside. Reading it with the entries on local LLMs and open source models gives the full picture of the options.
The name comes from the animal, a play on the LLM abbreviation, and it started a trend of animal names among open models.