Llama
Llama is Meta's family of open-weight large language models · first released in February 2023 · widely fine-tuned and self-hosted.
Text reviewed October 5, 2026
Llama is Meta's family of open-weight large language models · first released in February 2023 · widely fine-tuned and self-hosted.
Basic
Meta released Llama 1 in February 2023 for research, then Llama 2 (July 2023), Llama 3 (April 2024), Llama 3.1 (July 2024, including a 405B model) and Llama 4 (April 2025, the first mixture-of-experts Llamas). The weights can be downloaded and run on your own hardware or through many hosting providers.
Deep
Llama weights are released under the Llama Community License, which allows commercial use with conditions; it is not an OSI-approved open-source license, so "open-weight" is the accurate term. Because the weights are public, the same Llama model is hosted by many providers at different prices, and many community fine-tunes are built on it.
Expert
Meta published architecture details for its Llama releases (for example grouped-query attention since Llama 2 70B and mixture-of-experts in Llama 4). Self-hosting cost depends on parameter count, quantization and serving stack rather than on a per-token list price.
Meta keeps shipping new Llama models; the live block on this page lists the newest ones tracked on BenchGecko with their release dates, prices and BenchGecko scores, so this definition never has to name a "latest" model.
Depending on why you're here
- ·Meta's free-to-download AI models
- ·Anyone can run them on their own computers
- ·Self-host or pick the cheapest provider for the same weights
- ·Check the license terms before commercial use
- ·See /family/llama for every tracked model
- ·Open weights shift value from the model to hosting and tooling
- ·Dated Meta figures are on /economy/company/meta
- ·Published architectures and weights make Llama a common research baseline
- ·Mixture-of-experts since Llama 4