Llama 4 Maverick

Meta's open-weight multimodal MoE you can run yourself.

Intelligence

AA Intelligence Index14

independent benchmark score from Artificial Analysis, not Timace's own judgment

Best for

  • Self-hosted deployments with data-privacy requirements

Not ideal for

  • Matching frontier closed models on hard reasoning

Specifications

Developer
Meta

confidence 100%

www.llama.com

Verified Jul 21, 2026 · curated

Context window (tokens)
1M tokens

confidence 75%

www.llama.com

Verified Jul 21, 2026 · curated

Modalities
text, image, code

confidence 85%

www.llama.com

Verified Jul 21, 2026 · curated

API available
Yes

confidence 90%

www.llama.com

Verified Jul 21, 2026 · curated

Open source
Yes

confidence 95%

www.llama.com

Verified Jul 21, 2026 · curated

Compare against

Side-by-side →