AI Features Features How it works Pricing FAQ Blog Tutorial Glossary About Us Agencies
modele-ia

Llama (Meta)

Llama is the open-source large language model family developed by Meta. With 1.2 billion cumulative downloads, Llama is the world's most adopted open-source LLM, powering Meta AI and thousands of third-party applications.

What is Llama?

Llama (Large Language Model Meta AI) is the open-source large language model family developed by Meta. Launched in February 2023, Llama has become the world's most adopted open-source model ecosystem, with 1.2 billion cumulative downloads by April 2025.

Llama Models

  • Llama 4 Scout — 17B active parameters, 16 experts, 109B total parameters, 10 million token context window
  • Llama 4 Maverick — 17B active parameters, 128 experts, 400B total parameters, outperforms GPT-4o and Gemini 2.0 Flash
  • Llama 4 Behemoth (upcoming) — 288B active parameters, ~2 trillion total parameters, outperforms GPT-4.5 and Claude Sonnet 3.7

All Llama 4 models are natively multimodal with early fusion of text and vision, pre-trained on 200+ languages.

Llama by the Numbers

  • 1.2 billion cumulative downloads (1 million/day)
  • 9% of enterprise AI production usage
  • Powers Meta AI (1+ billion users)
  • Thousands of third-party applications built on Llama

Llama and AI Visibility

Llama is the engine behind Meta AI, but it also powers many third-party applications. An AI visibility audit must test Llama models as they are ubiquitous — from enterprise chatbots to personal assistants. AI Labs Audit includes Llama models in its a wide selection of tested models. Discover our tutorial.