Models · chat and writing, image understanding
Llama 4 Scout 109B-A17B: VRAM requirements and GPUs that run it
Meta MoE with 17B active and 10M-token context claims. Needs 64 GB+. At Q4 with 8K context it needs about 73 GB, so you want a GPU with at least 96 GB of VRAM or a Mac with 96 GB of unified memory.
Memory needed by quantization
| Quant | Weights | Total at 8K | Total at 32K |
|---|---|---|---|
| Q4_K_M | 71 GB | 73 GB | 77 GB |
| Q5_K_M | 85 GB | 87 GB | 92 GB |
| Q8_0 | 127 GB | 129 GB | 134 GB |
| FP16 | 240 GB | 242 GB | 247 GB |
Totals include about 10% runtime overhead and the KV cache, plus around 0.6 GB for the vision encoder. Longer conversations cost real memory: going from 8K to 32K context adds about 4.6 GB at any quantization. It is a mixture-of-experts model: 109B parameters in total but only about 17B are used for each token, so it runs much faster than a dense model of the same size and stays usable even when part of it sits in system RAM. Most people should download Q4_K_M; Q5 or Q8 only make sense if you have spare memory. Read which quantization to pick.
Which GPUs can run Llama 4 Scout 109B-A17B?
No single consumer GPU we track runs it fully in VRAM at 8K context.
Running it on a Mac
Apple Silicon Macs can give a model about 72% of their unified memory. Llama 4 Scout 109B-A17B fits from 96 GB (Apple M3 Max, about 14 tokens per second). With a faster chip like the Apple M3 Ultra (96 GB) expect around 28 tokens per second. Compare the options in our Mac vs NVIDIA guide.
Similar size models: Llama 3.3 70B, DeepSeek-R1 Distill 70B, gpt-oss 120B.
Compare Llama 4 Scout 109B-A17B
Questions
How much VRAM does Llama 4 Scout 109B-A17B need?
About 73 GB at Q4_K_M with 8K context, 129 GB at Q8_0, and 242 GB at full FP16. In practice that means a card with at least 96 GB of VRAM for Q4.
Can Llama 4 Scout 109B-A17B run on an 8 GB GPU?
No. It needs about 73 GB at Q4, well beyond 8 GB even with offload.
Can I run Llama 4 Scout 109B-A17B on a Mac?
Yes. The smallest configuration that runs it fully is Apple M3 Max with 96 GB of unified memory, at about 14 tokens per second. More memory bandwidth (Pro, Max, Ultra chips) means faster replies.
Estimates based on model size, architecture and memory bandwidth. Real numbers vary with software, drivers and context length.