Meta's open-weight Mixture-of-Experts flagship — 400B params, 1M token context, multimodal.
Llama 4 Maverick is Meta's most capable openly-licensed model — a sparse MoE that rivals closed frontier models while staying self-hostable.
Who it's for: Builders who want frontier-level quality with the freedom to self-host, fine-tune, and avoid vendor lock-in.
Activates a fraction of params per token — frontier quality at lower cost.
Ingest entire codebases and books in one window.
Reason over images, not just language.
Download, fine-tune, and deploy anywhere.
Llama 4 Maverick is the open-weight model to beat in 2026 — if you need frontier quality without a closed API, this is it.