Menu

Ollama hosting compared: GPU servers, CPU VPS and templates

We may earn commissions when you shop through the links below. Learn more

Ollama runs open-weight models behind a local HTTP API. Compare GPU and CPU hosts with official templates or documentation for Ollama, and size the GPU by the memory your model needs.

Stale · 8 providers · prices checked How we compare

Ollama hosting providers

8 of 8 providers · sorted by lowest listed price per GPU-hour (any GPU)

List prices from each provider's official pricing page, excluding VAT. Ordering and filters never depend on affiliate relationships. See the methodology.

Prices in USD, excluding VAT. ≈ marks a price the provider lists in EUR, converted at the ECB rate of 25 Sep 2026.≈ Estimated conversion from the listed USD price at the ECB reference rate of 23 Oct 2026 (1 EUR = 1.1193 USD). Billing currency not verified: providers may bill in USD, and your bank's rate and fees differ. Prices exclude VAT. EUR list prices are shown as the provider lists them (the dataset converts them to USD at the ECB rate of 25 Sep 2026).

Not sure Ollama hosting fits? Help me choose

What the official Ollama docs say

  • The Ollama server binds 127.0.0.1 on port 11434 by default. The OLLAMA_HOST environment variable changes the bind address.

    Source: Ollama docs: FAQ (opens in a new tab) · accessed

  • Ollama supports NVIDIA GPUs with compute capability 5.0 or higher and driver version 550 or newer; GPUs with compute capability 5.0 to 6.2 need driver 570 or newer.

    Source: Ollama docs: Hardware support (opens in a new tab) · accessed

  • Ollama Cloud is Ollama's own hosted service, separate from self-hosting. Ollama hosts those models and compute primarily in the United States and may route to Europe and Singapore for capacity.

    Source: Ollama pricing (opens in a new tab) · accessed

Ways to host: Ollama hosting

Compare by framework

  • Hermes Agent

    Open-source agent from Nous Research

    Hosts with an official template or docs: DigitalOcean, Fly.io, OVHcloud

    Compare Hermes Agent hosting

  • OpenClaw

    Always-on agent gateway, formerly Clawdbot

    Hosts with an official template or docs: DigitalOcean, Fly.io, Hetzner, Hyperstack, OVHcloud

    Compare OpenClaw hosting

  • Ollama

    Local models behind an HTTP API

    Hosts with an official template or docs: DigitalOcean, Hetzner, Hyperstack, Lambda, Modal, Nebius, OVHcloud, Scaleway

    You are on this page

  • vLLM

    High-throughput inference server

    Hosts with an official template or docs: DigitalOcean, Hyperstack, Lambda, Modal, Nebius, OVHcloud

    Compare vLLM hosting

  • Dify

    Open-source LLM app and workflow builder

    No host with an official template or docs in our data yet

    Compare managed hosting for Dify

  • Flowise

    Low-code builder for LLM flows and agents

    Hosts with an official template or docs: DigitalOcean, Nebius

    Compare managed hosting for Flowise

Tools for this decision

Explore related hubs

  • vLLM hosting6 providers

    GPU providers with vLLM templates or documentation.

  • Inference hosting9 providers

    Serverless and dedicated GPUs for serving models.

  • AI hosting10 providers

    Every provider in the dataset, across all AI hosting categories.

  • RTX 4090 cloudNo verified providers yet

    NVIDIA GeForce RTX 4090 24 GB for AI workloads.

  • L40S hosting5 providers

    NVIDIA L40S 48 GB for inference and fine-tuning.

  • AI VPS4 providers

    Virtual servers with root access for agents and automation.

Related Ollama guides

How we compare

  1. Prices per GPU-hour.

    GPU prices are USD per single GPU per hour from the provider's own pricing page or API. Multi-GPU nodes are divided by their GPU count, and a price only rented as a node says so (“rented as 8 GPUs”). Rows lead with the cheapest pay-as-you-go price (on-demand, secure, community or serverless). A spot, reserved or marketplace price leads only when the provider has no pay-as-you-go price here, or when you filter by that billing; it is labelled with its price type and listed in its own group after pay-as-you-go prices, never ranked against them. Prices exclude VAT and sales tax.

  2. Not verified means not verified.

    When no official source confirms a value we show “Not verified” instead of an estimate. Where a provider's terms do not allow its prices to be republished, we link to its live pricing instead.

  3. Freshness on every row.

    Updated today: checked on the current day (UTC). Fresh: up to 7 days. Stale: up to 30 days. Needs refresh: older than that.

  4. No paid placement.

    Results only change order with the sort you pick. Affiliate relationships never affect order, filters or which providers appear.

Filters match a provider when one of its listed offers meets every GPU, VRAM, region, price and billing-increment condition together. Spot, reserved and serverless billing list each provider that offers that pricing and show only prices on that basis, or “not verified” when we have none. Read the full methodology and our affiliate disclosure.

Frequently asked questions

How much VRAM do I need for Ollama?

Roughly the size of the quantised model weights plus room for the context window. The VRAM / Model Fit tool estimates this for a specific model and precision.

How do I run Ollama as a server?

Install it on a GPU server or VPS; it serves an HTTP API on port 11434, bound to 127.0.0.1 by default. To reach it remotely, keep it behind a VPN, an SSH tunnel or an authenticating reverse proxy rather than opening it with OLLAMA_HOST to the internet.

Is Ollama Cloud the same as hosting Ollama yourself?

No. Ollama Cloud runs models on Ollama's own infrastructure, which Ollama says is primarily in the United States (ollama.com/pricing, accessed 27 Sep 2026). Self-hosting means running Ollama on a server or GPU you rent, which is what this page compares.

Can Ollama run without a GPU?

Yes, small quantised models run on CPU, but token generation is much slower than on a GPU. Larger models need a GPU with enough VRAM.