Overview
Released in July 2024, Llama 3.1 405B was the largest openly available language model at the time and Meta’s first open model it described as competitive with leading closed models. It came alongside 8B and 70B versions sharing a 128K-token context window and support for eight languages.
Highlights
- 405 billion parameters, dense transformer architecture
- 128K-token context window
- Multilingual support for eight languages
- Tool use and long-form reasoning improvements over Llama 3
- Released with smaller 8B and 70B siblings and a detailed technical report
Typical uses
Self-hosted assistants for organizations that need data control, synthetic data generation, distillation into smaller models, and research.
Things to consider
- Running the full model requires multiple high-end GPUs; most users access it through cloud providers.
- The Llama license includes conditions, such as special terms for very large services and an acceptable use policy.
- Newer models have since surpassed it on many benchmarks.
Specifications are taken from the provider’s official announcement, documentation or model card. Benchmark results depend on evaluation settings; see how AI benchmarks work.