free open weights Engineers
Llama 4
by Meta
Meta's open-weight model family, a leading choice for self-hosted and customizable LLM deployments.
4/5
Visit Llama 4 - Context window
- Up to 10M tokens
- Parameters
- Mixture-of-experts, up to 400B
- Open weights
- Yes
- Released
- Apr 2025
Pros
- + Open weights — self-hostable and customizable
- + Very large context in some variants
- + Huge ecosystem and fine-tune support
Cons
- − Large variants need serious GPU hardware
- − Trails closed frontier models on hard reasoning
- − License has use restrictions
What Is Llama 4
Llama 4 is Meta’s open-weight large language model family, built on a mixture-of-experts architecture. It is the most widely deployed open model line in the industry.
In Practice
Because the weights are downloadable, Llama 4 is the default for teams that need on-premises deployment, data privacy, or deep customization through fine-tuning. It anchors a vast ecosystem of tools, quantizations, and derivative models.