Meta Releases Llama 3.1 405B Under Permissive License
Meta published the full 405 billion parameter Llama 3.1 weights on July 23, 2024. Anyone can download the checkpoint from Hugging Face, run inference on their own GPUs, or fine tune the model with LoRA adapters. Commercial redistribution and derivative products are explicitly allowed.
Teams no longer need to budget for closed API calls when prototyping large language model features. Budgets shift from inference spend to hardware or fine tuning time. Developers gain full control over data handling and model behavior.
Hugging Face hosts the official repository and reports more than 2 million downloads of the 405B weights in the first month. Perplexity AI has already fine tuned a Llama 3.1 405B variant for their search product and published benchmark gains of 7 percent on HotpotQA.
Step 1: Visit https://huggingface.co/meta-llama/Meta-Llama-3.1-405B and accept the license. Step 2: Run the command huggingface-cli download meta-llama/Meta-Llama-3.1-405B to fetch the weights. Step 3: Load the model in vLLM with python -m vllm.entrypoints.openai.api_server to obtain an OpenAI compatible endpoint for local testing.