Forex VPS built for terminals, at a price that holds on renewal. General VPS deployed in about 30 seconds.
By budget
Platforms
Brokers & prop firms
Toolsfree on every plan, add-ons switch on later
Compute
Dedicated Server
The whole machineNo neighbours, no shared I/O, root from the first minute.9machines7cities$99+a monthAvailable inDevelopers & DevOps
AI & Machine Learning
Pick the city closest to your broker’s server. Latency shown is node to broker hub.
Europe
London0.8 msAmsterdam0.8 msFrankfurt1 msZurich1 msParis1 msWarsaw3 msHelsinki3 msLimassol+ 4 more in EuropeAmericas
New York1 msChicago2 msMiami2 msToronto2 msDallas4 msLos Angeles4 msSão Paulo3 msAshburnWashington DC+ 4 more in the AmericasAsia-Pacific
Singapore0.8 msTokyo1 msHong Kong1 msSeoul2 msSydney2 msMumbai2 msNew Delhi3 ms+ 4 more in Asia-PacificRun the checker for a real figure from your own connection, not a guess from your timezone.
Check my latencyDomains
Web hosting
By what you run
Bundle
Register it here and it points at your hosting with nothing to configure.
Find a domainThe machine
Free with every plan
Billing and trust
See how to get one
Read moreAdd-onsswitch any of them on later, on any plan
Work with us
What you earn
Fund a balance, get up to 35% off every plan, and keep the difference. Our name appears nowhere.
See your marginWhat we hand youfree, and we keep them working as WHMCS and WordPress move
Nothing matched. Try a plan name or a city.
Ollama VPS
Ollama installs in one command and pulls its own weights. The question is not whether it runs — it is how fast, and that depends entirely on whether a card is doing the work. Both answers are on this page.
See the plans Which distribution?
From $6.37/mo on a yearly cycle · 2 GB holds a small model
Ollama itself is small. What fills the machine is the weights you keep loaded — a 3B model wants two gigabytes, a 7B wants eight.
STARTER-LNX
$2.79/ month
$33.50 billed yearly
Windows costs $4.50 for this shape
BASIC-LNX
$5.99/ month
$71.90 billed yearly
Windows costs $7.99 for this shape
TURBO-LNX
$8.39/ month
$100.70 billed yearly
Windows costs $11.99 for this shape
BOOST-LNX
$24.39/ month
$292.70 billed yearly
Windows costs $39.99 for this shape
ROYAL-LNX
$33.19/ month
$398.30 billed yearly
Windows costs $46.99 for this shape
SUPER-LNX
$42.79/ month
$513.50 billed yearly
Windows costs $63.99 for this shape
Every plan here undercuts its Windows counterpart — because there is no Microsoft licence inside it. That is the whole difference: same hardware, same cities, same support.
Any of them, on any plan, reinstallable from your panel in about five minutes.
Order to root
Nothing is approved by hand and nothing is queued. The only slow part is you deciding which city.
Two gigabytes runs a 3B model, eight runs a 7B. Size on the model you intend to keep loaded, not on Ollama itself.
0:00We never set a root password unless you ask. The key is safer and takes the same ten seconds.
0:01Ninety seconds on a KVM guest with its own kernel. Nothing to approve, nobody to wait for.
0:02One command installs it, one pulls a model. Bind the API to localhost and tunnel in, or put nginx and a token in front of it.
0:03Nothing is preinstalled. A clean distribution and root is easier to trust, and easier to rebuild, than an image we have already changed.
In your panel, not in a ticket queue
Buttons you press yourself, included on every plan — this one too. There are more than twenty in the panel; these are the ones that change the day.
One command, then it is yours
The install script sets up the service and the model store. No API key, no account, no per-token meter running in the background.
The weights stay on your disk
A pulled model sits in /usr/share/ollama and never leaves. For anything touching customer text that is usually the entire argument.
An API on port 11434
The same OpenAI-shaped endpoint your code already speaks. Point an app at it and nothing else changes.
Models you choose
Llama, Mistral, Qwen, Gemma, Phi. Pull one, delete it, pull another — the library is theirs and the disk is yours.
It stays loaded
Keep a model resident between requests rather than paying a cold start every time, which is what makes a small CPU model usable at all.
Move up when you need to
The same install runs on a GPU machine. When CPU speed stops being enough, the setup comes with you.
Twenty more sit alongside them — reboots, graphs, hostname changes, Windows Update control, time zone sync. The cheapest machine has the same panel as the most expensive. See how each one works →
The questions our support desk actually receives, answered plainly.
Still unsure which plan or which city? Send us your broker's account server name and we will measure it before you buy.
Real clients · since 2012
Here is somebody else's — and the node each of them actually runs on, so you can find the one closest to your own setup.
Latency around LD4 and Amsterdam is excellent — orders feel instant. We moved four accounts across and none of them noticed a thing.
Stable connection and quick help every time I have asked. Being next to the London hub is the whole reason my grid EA behaves now.
Excellent uptime and support is fast. It just works, which is all I wanted — I stopped thinking about the machine months ago.
Clear pricing and low latency. Migration was seamless — they moved the whole thing over a weekend and the renewal price has not moved since.
The Amsterdam route feels snappy and support solved my request quickly. Four stars only because I would like a German node closer to my broker.
Tokyo and Singapore peers are perfect for my brokers. Great value for what it is — I have run the same box for two years without touching it.
A person, not a queue. The same small team since 2012.
Only the essential ones run until you choose. Analytics and advertising wait for your yes. Cookie policy
Sign-in, your basket, security and this cookie choice itself. The site cannot work without them, so they cannot be turned off.
You can change this whenever you like — there is a Cookie settings link in the footer. See the cookie policy and the privacy policy.