Llama 3.2 Vision
First multimodal Llama models: 11B and 90B image-understanding releases
About
Llama 3.2 Vision, released September 25, 2024, was Meta's first Llama generation to accept images: an 11B and a 90B model pairing Llama 3.1 text backbones with vision adapters and a 128K context window. Meta reports the 90B reaching 60.3 on MMMU and 90.1 on DocVQA, roughly in the range of GPT-4o-mini on image tasks. It gave the open community a credible multimodal base to fine-tune for medical imaging, inspection, and document work without a closed API. The catch is the Llama 3.2 Community License, a custom commercial agreement with use restrictions, not a standard open-source license.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
MMMU 50.7/60.3, MathVista 51.5/57.3, DocVQA 88.4/90.1 for 11B/90B (Meta-reported)
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Release date 2024-09-25, 11B/90B sizes, 128K context verified in Meta's official model card; MMMU 50.7/60.3, MathVista 51.5/57.3, DocVQA 88.4/90.1 are Meta-reported; labeled as such; Llama 3.2 Community License is a custom commercial agreement, not an open-source license; caveat captured.