Mycelis
Meta
Llama 3.2 90B Vision Instruct is an open-source vision-language model from Meta. Focus areas: Vision, Reasoning, Enterprise. Context window: 32,768 tokens.
Context window
32.8k
Reasoning level
7 / 10
MMLU
84.0
HumanEval
78.0
Recommended deployment
GPU
A100 SXM
VRAM required
70 GB
Estimated cost
$2.35 / hour
Estimated from current GPU pricing, billed per hour while the deployment runs and stopped automatically when idle. The exact price depends on live GPU availability at deploy time.