What GPU do I need to run meta-llama/Llama-3.2-90B-Vision-Instruct?
88.6B parameters, published in BF16. View on Hugging FaceGated
Set up Llama-3.2-9…-Vision-InstructLlama-3.2-90B-Vision-Instruct is published by meta-llama on Hugging Face, with 103,812 downloads and 358 likes to date. It's a MllamaForConditionalGeneration model built for image-text-to-text, published natively in BF16, and gated: you'll need to accept the model's terms on Hugging Face before downloading weights.
VRAM required & cheapest live GPU fit
Required VRAM = weight size at each precision, plus a fixed overhead for KV-cache, activations, and fragmentation. Full formula and assumptions: methodology.
| Precision | Weight size | Required VRAM | Cheapest live fit | GPUs needed | Est. $/hr (full fit) |
|---|---|---|---|---|---|
| BF16 | 165.0 GB | 198.0 GB | RTX A6000 | 5 | $1.81/hr |
| FP8 (quantized) | 82.5 GB | 99.0 GB | RTX 5060 Ti | 7 | $0.770/hr |
| INT4 (quantized) | 41.3 GB | 49.5 GB | A100 | 1 | $1.21/hr |
| cheaper alt. | RTX 5060 Ti | 4 | $0.440/hr |
A GPU is only matched to a row if its hardware supports that precision, and the primary recommendation is always a single-GPU fit when one exists.
INT4 requires a quantized checkpoint actually published for this model, check its Hugging Face page before relying on this row.
Cheapest way to run Llama-3.2-90B-Vision-Instruct at its published (BF16) precision: 5× RTX A6000, at $0.363/hr per GPU ($1.81/hr total). Quantizing to FP8 or INT4 (rows above) can cost less, but requires a compatible quantized checkpoint to exist for this model.
Llama-3.2-90B-Vision-Instruct: common questions
Can Llama-3.2-90B-Vision-Instruct run on a single GPU?
No. At BF16 it needs 198.0 GB of VRAM, more than a single 32 GB desktop card holds. The cheapest capable card in the live feed is a 48.0 GB RTX A6000, and it takes 5 of them.
Do I need approval to download Llama-3.2-90B-Vision-Instruct?
Yes. meta-llama gates this repository on Hugging Face, so you have to accept its terms with a signed-in account before the weights will download. It does not change the 198.0 GB the model needs once you have them.
How many GPUs do I need to run Llama-3.2-90B-Vision-Instruct?
5 at BF16. It needs 198.0 GB of VRAM and the cheapest capable live offer is a 48.0 GB RTX A6000, so 5 of them come to $1.81/hr in total.
Does quantizing Llama-3.2-90B-Vision-Instruct lower the GPU bill?
Yes. At BF16 the cheapest live fit is 5 RTX A6000 cards at $1.81/hr. At FP8 (quantized) it drops to 7 RTX 5060 Ti cards at $0.770/hr, provided a quantized checkpoint exists for it.
Weight-to-VRAM math, the fit rules, and how live prices are normalized: full methodology.
More Llama 3.2 models
- Llama-3.2-11B-Vision-Instruct (10.7B, BF16)
- orpheus-3b-0.1-ft (3.8B, F32)
- svara-tts-v1 (3.3B, F32)
- orpheus-3b-0.1-ft (3.3B, BF16)
- alloma-3B-Instruct (3.2B, BF16)
Related reading: RTX A6000 pricing and specs, The NVIDIA Inception program, explained, Google Colab alternatives for dedicated GPU access, Free GPU credits for students and researchers, and RunPod volume disk vs. network volume, compared.
