Llama (Meta)
Llama is Meta's family of open-source large language models. Released with permissive licences, they can be downloaded, fine-tuned, and deployed on your own infrastructure — giving organisations full control over their AI stack without relying on a third-party API.
What is Llama?
Llama (Large Language Model Meta AI) is a series of open-source language models developed and released by Meta. The first version appeared in early 2023; Llama 2 followed later that year with a more permissive licence for commercial use; Llama 3, released in 2024, significantly improved performance across reasoning, instruction-following, and code tasks.
Unlike GPT-4 or Claude — which are accessible only via API, running on infrastructure controlled by OpenAI and Anthropic respectively — Llama models are openly distributed. You can download the weights, run them on your own servers or cloud environment, and fine-tune them on your own data without routing information through a third-party service.
Why Open-source Models Matter
The practical implications of open weights:
Data privacy: Documents processed on your own infrastructure never leave your environment. Relevant for supplier contracts, financial data, and customer records that cannot be sent to external APIs under data processing agreements.
Cost at scale: API costs compound at high volumes. A self-hosted Llama model has fixed infrastructure costs regardless of document volume.
Customisation: Fine-tuning on proprietary data is easier and cheaper when you control the weights.
Offline operation: Models can run in air-gapped environments with no internet connectivity.
Llama in Operations
For midsize manufacturers and wholesalers evaluating AI deployment options, Llama represents a credible alternative to cloud API-only approaches — particularly where data privacy concerns limit what can be sent externally, or where processing volumes make per-token API costs uneconomic. The trade-off is infrastructure overhead: you need GPU capacity and engineering capability to run and maintain the models. For organisations without that in-house, a managed deployment of Llama via a cloud provider (AWS Bedrock, Azure, Google Cloud) offers a middle path: open weights, managed infrastructure.