Plain-English Summary
Meta’s large open vision-language model for understanding images and documents.
Llama 3.2 90B Vision Instruct
Meta’s large open vision-language model for understanding images and documents.
Llama 3.2
Multimodal LLM
Open weights
Open weights: Yes
API: Via Meta and partners
Technical Notes
90B multimodal instruction model with image and text input, text output, and 128K context.
Best For
Document vision, image understanding, self-hosted multimodal RAG
Watch Out For
Large hardware requirement; primarily understanding rather than image generation.
Model Specs
- Maker
- Meta
- Country
- USA
- Family
- Llama 3.2
- Type
- Multimodal LLM
- Release date
- 2024-09-25
- Date confidence
- Exact
- Status
- Open weights
- Context window
- 131,072 tokens
- Max output
- Not listed
- Parameters
- 90B
- Active parameters
- 90B
- Architecture
- Vision-language transformer
- Reasoning
- General reasoning
- Tool calling
- Host dependent
- Structured output
- Host dependent
- API available
- Via Meta and partners
- Open weights
- Yes
- Self-hostable
- Yes
- License
- Llama 3.2 Community License
- Inputs
- Text, image
- Outputs
- Text
- Approx. price
- See provider pricing
- How to access
- Provider website
- Model / API ID
- Llama-3.2-90B-Vision-Instruct
- Knowledge cutoff
- Not publicly disclosed
- Fine-tuning
- Fine-tuning, quantization, self-hosting
- Completeness
- 95 %
- Verified
- Jul 18, 2026
- Snapshot date
- Jul 18, 2026
- Source
- Official source