← All models

Llama 3.2 90B Vision

Meta · 2024-09-25 · 88.6B parameters

Llama 3.2 90B Vision is Meta's pretrained base multimodal model from the September 2024 Llama 3.2 release. It pairs a Llama 3.1 text backbone with a separately trained image encoder connected through cross-attention adapter layers, takes text and images as input and generates text. It is the base checkpoint for Llama 3.2 90B Vision Instruct.

Benchmark scores