Llama 3.2 Vision 90B — VRAM, speed & local setup
Written by Jakub Rusinowski · Last updated
Full-power multimodal Llama. Requires high-end hardware (48+ GB VRAM) but matches GPT-4V on most vision benchmarks.
Llama 3.2 Vision 90B needs about 54 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache. On Apple Silicon that figure comes out of unified memory.
VRAM and speed by quantization
Quoted against NVIDIA RTX 4090 (24 GB). Includes the KV cache at 8K context, so it reads higher than the headline figure.
| Quant | Memory | VRAM | Speed (est.) | Fit |
|---|---|---|---|---|
| Q2_K 2.63 bpw | 33.3 GB | ~3 tok/s | Offload | |
| Q3_K_M 3.41 bpw | 42 GB | ~3 tok/s | Offload | |
| Q4_K_M 4.83 bpw | 57.8 GB | — | Too big | |
| Q5_K_M 5.67 bpw | 67.1 GB | — | Too big | |
| Q6_K 6.56 bpw | 77 GB | — | Too big | |
| Q8_0 8.50 bpw | 98.5 GB | — | Too big | |
| F16 16.00 bpw | 181.8 GB | — | Too big |
Black marker = usable memory on the NVIDIA RTX 4090 (24 GB). Estimates from the memory-bandwidth roofline on the methodology page. Llama 3.2 Vision 90B VRAM calculator →
Get Llama 3.2 Vision 90B running
The cheapest catalogued GPU that runs Llama 3.2 Vision 90B is the NVIDIA DGX Spark (128 GB).
or compare on Vast.ai from $0.77/hr (typical low · varies)
As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.
How to run Llama 3.2 Vision 90B
Install Ollama, then run:
ollama run llama3.2-vision:90bSpecifications
Curated — A hand-written entry from before this catalogue recorded its sources. The figures are long-standing but their provenance is not on file.
- Parameters
- 90 Billion
- Context window
- 128,000
- Architecture
- Dense + Vision Encoder
- Provider
- Meta
- Licence
- Llama 3.2 Community
- Practical deployment
- Workstation (large unified memory or pro GPU)
- Specified at
- Q4_K_M
- System RAM
- 64 GB
- Record updated
- 2024-09-25
Weights are downloadable and commercial use is permitted, subject to the licence’s acceptable-use terms.
Quality and use cases
Scores as published by the model’s authors or an independent evaluator — quality, not throughput, and not measured by us.
Can I run Llama 3.2 Vision 90B on my GPU?
- Llama 3.2 Vision 90B on AMD Radeon RX 9060 XT 8GB
- Llama 3.2 Vision 90B on AMD Ryzen AI Max+ 395
- Llama 3.2 Vision 90B on Apple M1 Ultra
- Llama 3.2 Vision 90B on Apple M2 Max
- Llama 3.2 Vision 90B on Apple M4 Max
- Llama 3.2 Vision 90B on Apple M5 Max
- Llama 3.2 Vision 90B on Intel Arc B570
- Llama 3.2 Vision 90B on Intel Arc B580
- Llama 3.2 Vision 90B on NVIDIA A100 80GB (PCIe)
- Llama 3.2 Vision 90B on NVIDIA DGX Spark
- Llama 3.2 Vision 90B on NVIDIA GeForce RTX 3060 (12GB)
- Llama 3.2 Vision 90B on NVIDIA GeForce RTX 3070
- Llama 3.2 Vision 90B on NVIDIA GeForce RTX 3070 Ti
- Llama 3.2 Vision 90B on NVIDIA GeForce RTX 3080 (10GB)
- Llama 3.2 Vision 90B on NVIDIA GeForce RTX 4060
- Llama 3.2 Vision 90B on NVIDIA GeForce RTX 4070
- Llama 3.2 Vision 90B on NVIDIA GeForce RTX 4070 Super
- Llama 3.2 Vision 90B on NVIDIA GeForce RTX 4070 Ti
- Llama 3.2 Vision 90B on NVIDIA GeForce RTX 5060
- Llama 3.2 Vision 90B on NVIDIA GeForce RTX 5060 Ti 8GB
- Llama 3.2 Vision 90B on NVIDIA GeForce RTX 5070
- Llama 3.2 Vision 90B on NVIDIA H100 80GB (PCIe)
- Llama 3.2 Vision 90B on NVIDIA RTX PRO 6000 Blackwell
Other Llama 3.2 Vision sizes
Llama 3.2 Vision 90B — frequently asked questions
How much VRAM does Llama 3.2 Vision 90B need?
About 54 GB at Q4_K_M — quantized weights plus framework overhead, before any KV cache. The cache grows with context length and is added on top; the table above folds it in. Apple Silicon counts unified memory toward the same figure.
Does Llama 3.2 Vision 90B run on an RTX 4090 (24 GB)?
No. Llama 3.2 Vision 90B needs about 54 GB at Q4_K_M, more than a single RTX 4090's 24 GB. It needs a larger card, several GPUs, or Apple Silicon with enough unified memory — or it runs with part of the weights offloaded to system RAM, which is much slower.
How do I run Llama 3.2 Vision 90B locally?
Install Ollama and run `ollama run llama3.2-vision:90b`. That pulls the weights and starts a local OpenAI-compatible endpoint; after the download nothing leaves the machine.
What other sizes does Llama 3.2 Vision come in?
Llama 3.2 Vision 11B (7 GB), Llama 3.2 Vision 90B (54 GB). Every size shares the family's training and licence; the larger ones score higher and need proportionally more memory.