How much VRAM does Llama 4 Scout (17B active / 109B total) need at Q4_K_M? Native 10M Context
Llama 4 Scout in resident mode at native 10M context requires massive datacenter setups (e.g., 28x H100 80GB cards). Active-only drops VRAM requirements slightly to ~2180 GB, which still demands an extreme server configuration.
Llama 4 Scout at Q4_K_M with its native 10M context needs about 2231 GB of VRAM with all 109B params resident - that's the number you size hardware against. Scout has 17B active parameters per token out of a 109B pool, 48 layers, 8 KV heads, and Meta's iRoPE interleaved rotary scheme that natively supports a 10M-token context window. As a single-node escape hatch, llama.cpp-style active-only loading drops the resident footprint to roughly 2180 GB at the cost of streaming cold experts through system RAM each token.
By TechCompare · Updated
Calculator
Estimated VRAM required
2335 GB
109B params at Q4_K_M, 10,485,760 token context, batch 1, inference.
Estimate accuracy: Weights within ~2%. KV cache within ~5% for standard GQA models, ~10% for MLA (DeepSeek). Real VRAM may vary with framework (vLLM vs llama.cpp vs Transformers), Flash Attention, and driver overhead.
KV cache exceeds model weights: Consider lowering the context length to save on VRAM. Contexts between 8K and 64K are generally more typical for local setups.
Hardware that fits
No single GPU in our catalog has enough memory. Multi-GPU or CPU offload required.
How this is calculated
The 109B parameter pool at Q4_K_M is 61 GB of weights, plus a massive 2062 GB KV cache (8 KV heads * 128 head_dim * 48 layers * 10M * 2 bytes) and ~109 GB overhead. That 2231 GB total is the resident-mode budget. Active-only loading keeps the same KV cache and activations but shrinks weights to 17B * 0.56 = 9.5 GB, totaling ~2180 GB - useful for fitting Scout on a pooled multi-GPU workstation or high-end unified memory setup.
Verdict
The cache math is what overwhelms Scout: 8 KV heads at head_dim 128 across 32 layers sums to ~1.66 GB per 1M tokens of context raw KV and the full 10M native window pushes the FP16 KV cache to around 2062 GB. Add 61 GB of Q4 weights and 109 GB overhead and the resident total lands at 2231 GB, which is 28x H100 80GB or 16x H200 141GB with NVLink. Active-only loading keeps the same KV cache (it's bound by context length, not expert routing), so dropping weights to 9.5 GB only trims the total to around 2180 GB. The right way to fit Scout on a single card is to cap context: at 32K, active-only drops to about 18 GB and lands cleanly on a 24 GB consumer GPU.
More Llama scenarios
Related guides
Frequently asked questions
What's iRoPE and why does it support 10M context?
Can Scout run on a single 24 GB GPU?
What about Maverick (17B/400B)?
Related tools
RAM Latency Calculator
Convert DDR3/DDR4/DDR5 timings (CL, tRCD, tRP, tRAS) into true latency in nanoseconds.
Use tool ➜Power Cost Estimator
Estimate annual electricity costs for your PC, Server, or TV.
Use tool ➜Data Transfer Calculator
Estimate transfer times for files over USB, WiFi, Ethernet, and more.
Use tool ➜Memory and Storage Latency Visualizer
Visualize the massive speed difference between CPU cache, RAM, and storage.
Use tool ➜