What is Meta AI? — AI Glossary

What it is: Meta AI is the AI division at Meta (Facebook, Instagram, WhatsApp). It builds the Llama family of open-weight large language models, the Meta AI assistant integrated across Facebook/Instagram/WhatsApp, and research projects like Ray-Ban Meta smart glasses.
Who it is for: Anyone curious about Meta’s AI strategy, or developers using Llama in their own projects.
Best if: You want to understand why Meta releasing Llama as open-weights reshaped the developer AI landscape, or you use the Meta AI assistant inside WhatsApp/Instagram.
Skip if: You don’t use Meta’s apps and aren’t a developer — Meta AI rarely surfaces outside the Meta ecosystem. Want one practical AI workflow every morning? Subscribe to our free daily newsletter.

What is Meta AI?

Meta AI is the artificial-intelligence research and product arm at Meta Platforms, led by Yann LeCun (Chief AI Scientist) and a research division that publishes some of the most-cited papers in the field. Internally, Meta refers to it as FAIR (Fundamental AI Research). Externally, “Meta AI” is the name of the assistant users interact with inside Facebook, Instagram, WhatsApp, and Ray-Ban Meta smart glasses.

Why does Meta AI matter?

Meta’s defining move was releasing the Llama family of large language models as open-weight, starting with Llama 1 in February 2023. This single decision arguably reshaped the entire open-source AI landscape: developers and smaller companies suddenly had access to capable models they could fine-tune, run privately, or embed in products without per-token API costs to OpenAI or Anthropic.

The Meta AI consumer assistant uses Llama under the hood and is available in 40+ countries inside Meta’s apps. Other notable Meta AI projects: Make-A-Video (text-to-video research), Audiobox (audio generation), the SeamlessM4T translation models, and an active research program in computer vision via Ray-Ban Meta smart glasses.

How does Meta AI compare to OpenAI and Google?

The key strategic difference: Meta open-sources its frontier models; OpenAI and Anthropic keep theirs closed. Google sits in the middle, releasing Gemma (open-weight) but keeping Gemini Pro closed.

This positioning lets Meta argue it’s democratizing AI access while also seeding adoption that benefits Meta’s broader platform strategy. Critics argue open-weight frontier models present greater misuse risks. Supporters counter that the open-weight community surfaces safety problems faster and prevents AI concentration in a handful of US companies. The debate continues; Meta has remained committed to open-weight releases through Llama 1, 2, 3, 3.1, 4, and beyond.

Related terms

• Llama
• Yann LeCun
• Open Weights
• Google DeepMind
• Large Language Model

Learn more on Beginners in AI

• Meta AI — Site Coverage

Sources and further reading

• Meta AI — official
• Llama models — official
• Meta AI Research

Last reviewed: May 2026. AI terminology evolves quickly — verify specifics on the official source pages above.

You may also like

• Meta AI — Site Coverage
• Llama
• Yann LeCun
• Open Weights
• Google DeepMind
• AI Glossary: 100+ Terms Every Beginner Needs to Know