Run Llama 4 Scout (109B MoE) on a VPS.
Prices checked
The cheapest VPS that runs Llama 4 Scout (109B MoE) is Leaseweb VPS 6 at $65.69 a month, with 120 GB of RAM for the 68.4 GB it needs, checked .
Best plans for Llama 4 Scout (109B MoE)
Most tokens per second for your money, comfortable fits first.
- 1
VPS 6 · 24 vCPU / 120 GB · US
Leaseweb · 24C · 120 GB · ~1.2–2.3 tok/s$65.69/mo - 2
VPS 6 · 24 vCPU / 120 GB · UK
Leaseweb · 24C · 120 GB · ~1.2–2.3 tok/s$69.93/mo - 3
VPS 6 · 24 vCPU / 120 GB · Europe
Leaseweb · 24C · 120 GB · ~1.2–2.3 tok/s$70.37/mo - 4
VPS 6 · 24 vCPU / 120 GB · Singapore
Leaseweb · 24C · 120 GB · ~1.2–2.3 tok/s$109.50/mo - 5$203.62/mo
- 6
VPS 6 · 24 vCPU / 120 GB · Australia
Leaseweb · 24C · 120 GB · ~1.2–2.3 tok/s$158.03/mo
netcup · 20C · 96 GB · ~1.7–3.5 tok/sWhat it needs
RAM needed
68.4 GB
CPU inference at Q4_K_M
Weights
65.4 GB
Q4_K_M
KV cache
1.5 GB
8K context
OS and runtime
1.5 GB
headroom
Mixture-of-experts: 109B of weights to hold in RAM but only 17B active per token, so it runs surprisingly fast once it fits.
- Size
- 109B total · 17B active
- Context
- 10.24M tokens
- Kind
- chat, vision
- Released
- 2025-04
- License
- Llama 4 Community License
Facts from Hugging Face: exact GGUF file sizes, KV geometry from the GGUF header, 128,495 downloads.
Or rent it as an API
every price for Llama 4 Scout20 providers sell Llama 4 Scout.
Meta's own price
$0.15per M tokens
Cheapest price
$0.06per M tokens
Lambda
What the VPS price buys
1051M tokens a month
at $65.69 a month
Or rent a GPU
every price for the A100 80GB PCIe2 providers rent the A100 80GB PCIe, with 80 GB of GPU memory against the 68.4 GB Llama 4 Scout (109B MoE) needs.
Cheapest board that fits
A100 80GB PCIe
80 GB of GPU memory
On demand
$1.19per GPU an hour
RunPod
Held for a month
$868.70
730 hours of A100 PCIe · Community Cloud
At 5 million tokens a month, what this box generates in 730 hours at its estimated rate, the API costs you $0.31, this VPS $65.69 and the GPU $868.70. Above about 1,051 million tokens a month the VPS costs less than the API, which is more than this box generates in a month.
Or buy hardware
all machines3 reference machines fit Llama 4 Scout (109B MoE).
Framework Desktop (Ryzen AI Max+ 395, 128 GB)
Approximate128 GB · 256 GB/s · ~6.5–10 tok/s
$3,449= 53 mo of VPSNVIDIA DGX Spark (128 GB)
Approximate128 GB · 273 GB/s · ~7–11 tok/s
$4,699= 72 mo of VPSMac Studio (M5 Ultra, 96 GB)
Tight fit96 GB · 1200 GB/s · ~31–47 tok/s
$5,499= 84 mo of VPS
Buy · per month
$100.47
$95.81 hardware + $4.67 power
Rent · per month
$65.69
Leaseweb VPS 6 · 24 vCPU / 120 GB · US
Break-even
The Framework Desktop (Ryzen AI Max+ 395, 128 GB) pays for itself after 57 months of replacing the VPS, and it streams an estimated 6.5–10 tok/s against the VPS's CPU-only pace.
How we count
On a CPU-only VPS Llama 4 Scout (109B MoE) needs about 68.4 GB of RAM: 65.4 GB of Q4_K_M weights, 1.5 GB of KV cache for an 8,192-token context and 1.5 GB for the OS and runtime. 9 of the 527 plans in the index fit and include a disk. A plan runs it comfortably with 20% headroom over the total, and is a tight fit with exactly enough.
Other quants: Q8_0 weights take 114.5 GB (near-lossless, about half of F16); F16 takes 1.8 GB.
The plans are ranked by estimated tokens per second per dollar, comfortable fits first. Speed is effective memory bandwidth divided by active weight bytes (4 GB/s per shared vCPU, 6 per dedicated, times 0.5 for MoE routing overhead), shown as a band. Real numbers depend on the host CPU generation, AVX-512 and AMX support and how noisy the neighbours are, so treat them as an order of magnitude.
Token prices are per million, blended 3 in : 1 out, before tax. A price below the maker's is listed only when the provider says why. A rented box costs the same whether or not you use it and runs whatever else you put on it; an API charges for what you send and nothing when you stop.
A GPU rate is per GPU for an hour, and the monthly figure is 730 hours of the whole instance at it, tax excluded. The token price is $0.06 per million, blended 3 in : 1 out, the cheapest price (Lambda), and the box is assumed busy every second of the month, which nothing is.
A machine tagged approximate is priced at the typical US street price in August 2026 rather than its maker's own price. Local speed is peak bandwidth times 0.7 efficiency (0.35 on CPU-only boards) divided by active weight bytes; unified-memory machines are assumed to give models 75% of their RAM. GPU cards need a host PC that is not in the price.
Frequently asked
- How much RAM does Llama 4 Scout (109B MoE) need?
- About 68.4 GB for CPU inference at Q4_K_M: 65.4 GB of weights, 1.5 GB of KV cache at 8,192 tokens of context, and 1.5 GB of headroom. Longer contexts need more KV cache (192 KB per token for this model).
- What is the cheapest VPS that can run Llama 4 Scout (109B MoE)?
- Leaseweb VPS 6 · 24 vCPU / 120 GB · US (120 GB RAM, 24 vCPU) at $65.69/mo excl. VAT runs it comfortably as of 31 Oct 2026.
- How fast will Llama 4 Scout (109B MoE) run on a VPS without a GPU?
- Roughly 1.2–2.3 tok/s on the top pick. Token generation is limited by memory bandwidth, because each new token reads the active experts' weights once. More vCPUs help, dedicated ones most, but a GPU or an Apple silicon machine is 10 to 50 times faster.
Same size class