Meta's open-weight 405B flagship — near-frontier quality you can self-host and fine-tune.
Llama 3.1 (405B) was Meta's breakthrough open model — the first openly licensed model to match closed frontier systems on many benchmarks. With 8B, 70B, and 405B sizes, it gives teams a self-hostable option for everything from chatbots to RAG pipelines. Because the weights are open, you can run it privately, fine-tune it, and avoid per-token API bills.
Who it's for: Teams that need frontier-grade quality with data control, self-hosting, or custom fine-tuning — and don't want vendor lock-in.
Download, self-host, and modify under Meta's community license.
Long enough for most documents and multi-turn agents.
8B for edge, 70B for servers, 405B for frontier tasks.
LoRA / full fine-tunes on your own data.
Llama 3.1 is the default open model for teams that want control. If you can run it, you get near-frontier quality without sending data to a third party.