Can I run Llama 3.2 11B Vision Instruct locally?
Llama 3.2 11B Vision Instruct by meta-llama has 11B parameters. At 4-bit with an 8K context it needs about 7.4 GB of GPU memory. It fits on 65 of 65 GPUs and Macs we track. See all models that fit in 8 GB of VRAM.
Memory needed by quant and context length
GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from leafspark/Llama-3.2-11B-Vision-Instruct-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.
| Quant | Quality | 4K ctx | 8K ctx | 32K ctx | 128K ctx |
|---|---|---|---|---|---|
| 2-bit (Q2_K) | rough | 5.0 GB | 5.6 GB | 9.4 GB | 24.4 GB |
| 3-bit (Q3_K_M) | usable | 6.1 GB | 6.8 GB | 10.5 GB | 25.5 GB |
| 4-bit (Q4_K_M) • | good, most popular | 6.8 GB | 7.4 GB | 11.2 GB | 26.2 GB |
| 5-bit (Q5_K_M) | very good | 8.4 GB | 9.0 GB | 12.7 GB | 27.7 GB |
| 6-bit (Q6_K) | near lossless | 9.4 GB | 10.0 GB | 13.8 GB | 28.8 GB |
| 8-bit (Q8_0) • | effectively lossless | 11.0 GB | 11.6 GB | 15.3 GB | 30.3 GB |
| 16-bit (F16/BF16) • | original weights | 19.5 GB | 20.1 GB | 23.9 GB | 38.9 GB |
Which GPUs and Macs run Llama 3.2 11B Vision Instruct?
Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).
| Hardware | Memory | Fits? | Best quant | Needs | Max tok/s (est.) | |
|---|---|---|---|---|---|---|
| RTX 3060 12GB | 12 GB | Fits | Q6_K | 10.0 GB | 41 | details |
| RTX 4060 Ti 16GB | 16 GB | Fits | Q8_0 | 11.6 GB | 28 | details |
| RTX 3090 | 24 GB | Fits | F16 | 20.1 GB | 48 | details |
| RTX 4090 | 24 GB | Fits | F16 | 20.1 GB | 52 | details |
| RTX 5090 | 32 GB | Fits | F16 | 20.1 GB | 92 | details |
| RTX 5070 Ti | 16 GB | Fits | Q8_0 | 11.6 GB | 86 | details |
| RX 7900 XTX | 24 GB | Fits | F16 | 20.1 GB | 49 | details |
| Mac M4 Pro 48GB | 48 GB | Fits | F16 | 20.1 GB | 14 | details |
| Mac M4 Max 128GB | 128 GB | Fits | F16 | 20.1 GB | 28 | details |
| Mac M3 Ultra 512GB | 512 GB | Fits | F16 | 20.1 GB | 42 | details |
| DGX Spark | 128 GB | Fits | F16 | 20.1 GB | 14 | details |
| RTX 4070 | 12 GB | Fits | Q6_K | 10.0 GB | 58 | details |
| RTX 5060 Ti 16GB | 16 GB | Fits | Q8_0 | 11.6 GB | 43 | details |
| RTX 4060 | 8 GB | Fits | Q3_K_M | 6.8 GB | 52 | details |
| Arc B580 | 12 GB | Fits | Q6_K | 10.0 GB | 52 | details |
| RTX 3060 Ti | 8 GB | Fits | Q3_K_M | 6.8 GB | 86 | details |
| RTX 3070 | 8 GB | Fits | Q3_K_M | 6.8 GB | 86 | details |
| RTX 3080 10GB | 10 GB | Fits | Q4_K_M | 7.4 GB | 127 | details |
| RTX 3080 12GB | 12 GB | Fits | Q6_K | 10.0 GB | 104 | details |
| RTX 3090 Ti | 24 GB | Fits | F16 | 20.1 GB | 52 | details |
| RTX 4060 Ti 8GB | 8 GB | Fits | Q3_K_M | 6.8 GB | 55 | details |
| RTX 4070 Super | 12 GB | Fits | Q6_K | 10.0 GB | 58 | details |
| RTX 4070 Ti | 12 GB | Fits | Q6_K | 10.0 GB | 58 | details |
| RTX 4070 Ti Super | 16 GB | Fits | Q8_0 | 11.6 GB | 65 | details |
| RTX 4080 | 16 GB | Fits | Q8_0 | 11.6 GB | 69 | details |
| RTX 4080 Super | 16 GB | Fits | Q8_0 | 11.6 GB | 71 | details |
| RTX 5060 | 8 GB | Fits | Q3_K_M | 6.8 GB | 86 | details |
| RTX 5070 | 12 GB | Fits | Q6_K | 10.0 GB | 77 | details |
| RTX 5080 | 16 GB | Fits | Q8_0 | 11.6 GB | 92 | details |
| 2x RTX 3090 | 2×24 GB | Fits | F16 | 20.1 GB | 48 | details |
| 2x RTX 4090 | 2×24 GB | Fits | F16 | 20.1 GB | 52 | details |
| 2x RTX 5090 | 2×32 GB | Fits | F16 | 20.1 GB | 92 | details |
| RTX A6000 | 48 GB | Fits | F16 | 20.1 GB | 39 | details |
| RTX 6000 Ada | 48 GB | Fits | F16 | 20.1 GB | 49 | details |
| RTX PRO 6000 Blackwell | 96 GB | Fits | F16 | 20.1 GB | 92 | details |
| L40S | 48 GB | Fits | F16 | 20.1 GB | 44 | details |
| A100 40GB | 40 GB | Fits | F16 | 20.1 GB | 80 | details |
| A100 80GB | 80 GB | Fits | F16 | 20.1 GB | 104 | details |
| H100 PCIe | 80 GB | Fits | F16 | 20.1 GB | 102 | details |
| H100 SXM | 80 GB | Fits | F16 | 20.1 GB | 171 | details |
| H200 | 141 GB | Fits | F16 | 20.1 GB | 245 | details |
| RX 7600 | 8 GB | Fits | Q3_K_M | 6.8 GB | 55 | details |
| RX 7700 XT | 12 GB | Fits | Q6_K | 10.0 GB | 49 | details |
| RX 7800 XT | 16 GB | Fits | Q8_0 | 11.6 GB | 60 | details |
| RX 7900 GRE | 16 GB | Fits | Q8_0 | 11.6 GB | 55 | details |
| RX 7900 XT | 20 GB | Fits | Q8_0 | 11.6 GB | 77 | details |
| RX 9070 | 16 GB | Fits | Q8_0 | 11.6 GB | 62 | details |
| RX 9070 XT | 16 GB | Fits | Q8_0 | 11.6 GB | 62 | details |
| Ryzen AI Max+ 395 128GB | 128 GB | Fits | F16 | 20.1 GB | 13 | details |
| Arc A770 16GB | 16 GB | Fits | Q8_0 | 11.6 GB | 54 | details |
| Mac M1 16GB | 16 GB | Fits | Q5_K_M | 9.0 GB | 8.9 | details |
| Mac M2 24GB | 24 GB | Fits | Q8_0 | 11.6 GB | 9.6 | details |
| Mac M3 24GB | 24 GB | Fits | Q8_0 | 11.6 GB | 9.6 | details |
| Mac M4 16GB | 16 GB | Fits | Q5_K_M | 9.0 GB | 16 | details |
| Mac M4 32GB | 32 GB | Fits | Q8_0 | 11.6 GB | 12 | details |
| Mac M1 Pro 32GB | 32 GB | Fits | Q8_0 | 11.6 GB | 19 | details |
| Mac M3 Pro 36GB | 36 GB | Fits | F16 | 20.1 GB | 7.7 | details |
| Mac M4 Pro 24GB | 24 GB | Fits | Q8_0 | 11.6 GB | 26 | details |
| Mac M4 Pro 64GB | 64 GB | Fits | F16 | 20.1 GB | 14 | details |
| Mac M1 Max 64GB | 64 GB | Fits | F16 | 20.1 GB | 20 | details |
| Mac M2 Max 96GB | 96 GB | Fits | F16 | 20.1 GB | 20 | details |
| Mac M3 Max 128GB | 128 GB | Fits | F16 | 20.1 GB | 20 | details |
| Mac M4 Max 64GB | 64 GB | Fits | F16 | 20.1 GB | 28 | details |
| Mac M2 Ultra 192GB | 192 GB | Fits | F16 | 20.1 GB | 41 | details |
| Mac M3 Ultra 256GB | 256 GB | Fits | F16 | 20.1 GB | 42 | details |
Questions
How much VRAM does Llama 3.2 11B Vision Instruct need?
About 7.4 GB at 4-bit (Q4_K_M) with an 8K context: 5.6 GB of weights (real GGUF file size), 1.3 GB of KV cache and 0.6 GB of runtime overhead. At 8-bit it needs about 11.6 GB.
What is the smallest GPU that runs Llama 3.2 11B Vision Instruct?
At Q4_K_M and 8K context, the smallest single consumer GPU in our table that fits it is the RTX 3080 10GB (10 GB).
Can I run Llama 3.2 11B Vision Instruct on a Mac?
Yes. At Q4_K_M with 8K context it fits on a Mac M1 16GB or larger, using the default macOS GPU memory limit.
Similar models
Same family or a similar parameter count, with the VRAM each needs at Q4_K_M and 8K context.
Llama 3.2 11B Vision
11B · 8.0 GB at 4-bitLlama Guard 4 12B
12B · 9.1 GB at 4-bitLlama 3.1 8B Instruct
8.0B · 6.2 GB at 4-bitGLM 4.6V Flash
10B · 6.7 GB at 4-bitGLM 4.1V 9B Thinking
10B · 6.7 GB at 4-bitStep3 VL 10B
10B · 6.5 GB at 4-bitModel facts
| Hugging Face repo | meta-llama/Llama-3.2-11B-Vision-Instruct |
| Parameters | 10,670,220,835 (11B) |
| Layers / hidden size | 40 / 4096 |
| Attention | GQA; 32 heads, 8 KV heads, head dim 128 |
| KV cache per token (FP16) | 160.0 KiB |
| Max context | 131,072 tokens |
| License | llama3.2 (gated: accept terms on Hugging Face) |
| Released on HF | Sep 18, 2024 |
| Downloads (30d) / likes | 51,416 / 1,698 |
Sources: Hugging Face model API · config.json (unsloth/Llama-3.2-11B-Vision-Instruct) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.