Rent a GPU

Llama 4 Scout 17B-16E hardware requirements

109B parameters (17B active, mixture of experts). Needs about 68.3 GB at Q4_K_M with 8K context.

Parameters109B17B active (MoE)
Q4_K_M · 8K68.3 GBthe usual download
FP16 · 8K220.2 GBfull precision
Runs well on4/51GPUs and Macs we track

Memory needed at each quantization

Weights, KV cache and runtime overhead, at 8K and 32K tokens of context.

QuantQuality8K context32K context
FP16Full quality220.2 GB225 GB
Q8_0Near-lossless118 GB122.9 GB
Q6_KExcellent91.6 GB96.4 GB
Q5_K_MVery good79.7 GB84.6 GB
Q4_K_MGood, most popular68.3 GB73.1 GB
Q3_K_MNoticeably degraded55.5 GB60.3 GB
Q2_KHeavily degraded43.1 GB47.9 GB

Run it with Ollama

Downloads Q4_K_M by default.

ollama run llama4:scout

Model card

meta-llama/Llama-4-Scout-17B-16E-Instruct on Hugging Face: licence, prompt format and GGUF builds.

Run it in the cloud

Run Llama 4 Scout 17B-16E on a rented GPU today.

It needs about 68.3 GB at Q4_K_M: a rented A100 / H100 80 GB runs it for about $1.11 to $3.99 an hour.

RunPodRent a GPU on RunPodVast.aiRent a GPU on Vast.ai

Every GPU and Mac checked

Sorted by memory. Bars show how much of each card Llama 4 Scout 17B-16E fills at the best quant that fits.

4 Runs well

NameMemory usedBest quantSpeed

3 Fits with heavy quantization

NameMemory usedBest quantSpeed

44 Too big

NameMemory usedBest quantSpeed

Other Llama models