Skip to content

AI VPS

A VPS to run your LLM with Ollama

Keep your prompts, documents and generated answers on French infrastructure: Ollama runs models such as Llama, Phi or Mistral, and Open WebUI adds a chat interface. Our VPSs have no GPU: they suit small quantised models and moderate use.

From €12.99 /month, no commitment

  • RAM for the model

    From 8 GB (VPS Plus) to 32 GB (VPS Titan): the model, its context and its cache must fit in memory without swapping.

  • CPU inference, no illusions

    No GPU: aim for Llama 3.2 3B, a compact Phi variant or Mistral 7B in Q4 with enough RAM.

  • NVMe for your models

    Each model weighs several gigabytes; the guide recommends at least 30 to 50 GB, and the VPS Plus provides 60.

  • Queries kept in France

    Your conversations stay on a VPS hosted in France, as long as you don't enable an external provider.

Recommended plans

Billing period
  • Best balance

    VPS Plus

    The balanced setup for your production apps and panels.

    €12.99 /moNo commitment

    Order
    • 4 dedicated vCPUs
    • 8 GB DDR4
    • 60 GB NVMe
    • 500 Mbps network
    • Full root access
    • 1 dedicated IPv4 address
    • noVNC console
    • Automatic backup
    • Firewall and reverse DNS
  • Demanding apps

    VPS Pro

    More memory for databases and resource-hungry applications.

    €18.99 /moNo commitment

    Order
    • 4 dedicated vCPUs
    • 12 GB DDR4
    • 80 GB NVMe
    • 500 Mbps network
    • Full root access
    • 1 dedicated IPv4 address
    • noVNC console
    • Automatic backup
    • Firewall and reverse DNS
  • Heavy workloads

    VPS Ultra

    For heavy workloads: lots of containers, CI, large websites.

    €38.99 /moNo commitment

    Order
    • 8 dedicated vCPUs
    • 24 GB DDR4
    • 150 GB NVMe
    • 1 Gbps network
    • Full root access
    • 1 dedicated IPv4 address
    • noVNC console
    • Automatic backup
    • Firewall and reverse DNS

Prices in euros, VAT not applicable (article 293 B of the French General Tax Code). Secure payment in the client area, automatic delivery.

See the full Linux VPS range →

French infrastructure, verifiable numbers

  • L3/L4Anti-DDoS included

    Network attack detection and mitigation on every plan, at no extra cost.

  • 3.5 GHzXeon CPU

    Intel Xeon Gold 6138 and E5-2673 v4, up to 3.5 GHz turbo for your game servers.

  • 100%NVMe storage

    No spinning disks. Worlds and databases that respond fast.

  • < 15 msto Paris

    Infrastructure hosted in France, close to Paris for a low ping.

Hardware, protection, backups and availability in detail →

Frequently asked questions

How much RAM does Ollama need on a VPS?

8 GB is a realistic starting point for a quantised model with 1 to 3 billion parameters (VPS Plus). 16 GB (VPS Elite) is more comfortable and allows some 7B models in Q4. Beyond that, extra RAM adds headroom without guaranteeing good performance on CPU.

Do your VPSs have a GPU?

No, TalCloud does not offer GPU VPSs. Inference runs on CPU: it suits prototyping, a lightly used internal assistant or a low-volume n8n integration, not a busy public chatbot.

Can I get a ChatGPT-style interface?

Yes, with Open WebUI: persistent conversations, user accounts, model selection and RAG over your documents. It installs as a Docker container alongside Ollama, behind an HTTPS reverse proxy.

Should I expose the Ollama API to the internet?

No. The API on port 11434 has no built-in authentication: keep it on 127.0.0.1 or a private Docker network, and go through a reverse proxy with HTTPS and access control if a remote application needs to call it.

Other Linux VPS use cases