What AI models does Meta AI use?

Question

Accepted Answer

Meta AI runs on Llama (Meta), Llama 3 (Meta), Llama 3.3 (Meta), Llama 4 (Meta), Llama 4 Maverick (Meta), Llama 4 Scout (Meta), Muse (Meta), Muse Glimmer (Meta), Muse Image (Meta), Muse Spark (Meta), Muse Spark 1.1 (Meta), Muse Spark 1.2 (Meta), Muse Spark 1.3 (Meta), Muse Spark Open (Meta), Muse Voice Transcribe (Meta), Muse Wonder (Meta), Muse Wonder Full (Meta), SAM 3.1 (Meta) and Segment Anything (SAM) 3.1 (Meta), and routes requests between them. Model claims are labelled by how we know them: Llama — inferred, Llama 3 — inferred, Llama 3.3 — disclosed, Llama 4 — disclosed, Llama 4 Maverick — disclosed, Llama 4 Scout — disclosed, Muse — inferred, Muse Glimmer — disclosed, Muse Image — disclosed, Muse Spark — disclosed, Muse Spark 1.1 — disclosed, Muse Spark 1.2 — disclosed, Muse Spark 1.3 — disclosed, Muse Spark Open — disclosed, Muse Voice Transcribe — disclosed, Muse Wonder — disclosed, Muse Wonder Full — disclosed, SAM 3.1 — disclosed and Segment Anything (SAM) 3.1 — disclosed. "Disclosed" means the vendor says so in public; "inferred" means it was derived from documentation, headers or release notes. Context window: 1M. Other stack signals: Llama, Meta MTIA and PyTorch.

Cite this answer