AIHostBox
Client area Deploy an app
All 57 applications
CategoriesAI assistants & chatLLM runtimesAgents & RAGGateways & LLM opsVector & dataAutomationSpeech, image & videoSearch & web agentsSupport & CXML platformMCP servers
SolutionsPrivate team chatUnlimited automationCompany knowledge AIAI gateway & observability
More PricingGuides ComparisonsToolsPrivate LLM hosting LicensingSecurity SupportClient area

Ollama hosting

Your private LLM server — pull a model, get an API, keep every token on your own box.

LLM runtimesMITfrom $16.99/mo
In short

Ollama hosting at AIHostBox means a dedicated, single-tenant Ollama instance on our infrastructure: we install it, pin it to a reviewed release, put it behind TLS on your domain, back it up nightly and keep it patched. Plans start at $16.99 per month and the admin account, the data and any API keys are yours.

OpenAI-compatible APILlamaQwenMistralGemmaModel library

Ollama turns open-weight models — Llama, Hermes, Qwen, Mistral, Gemma and hundreds more — into a one-line experience: ollama pull, and you have a private, OpenAI-compatible API. Nothing is metered, no prompt leaves the machine, and no provider decides what your model may answer.

Our plans run Ollama on dedicated CPU cores with NVMe storage for model files. To be straight with you: these are CPU plans. Quantized 7-8B models are genuinely usable for chat, RAG and automation glue at roughly 5-15 tokens per second; 13B works on the 12 GB plan for non-realtime jobs. If you need a 70B model or realtime latency you want GPU hardware, and we would rather say so here than after checkout.

Ollama hosting plans

StarterRuns 7-8B models (Q4)
8 GB RAM · 4 vCPU · NVMe
$16.99/mo
Order
PlusRuns 13B models (Q4)
12 GB RAM · 6 vCPU · NVMe
$27.99/mo
Order

Every plan is a dedicated single-tenant instance with NVMe storage, nightly off-box backups, TLS and managed patching. Quarterly, semi-annual and annual terms are discounted at checkout. Model tokens are not included — you bring your own key.

How we run Ollama

The official ollama/ollama container serving its API on port 11434, with model files on a persistent NVMe volume at /root/.ollama and a TLS reverse proxy with key authentication in front.

What we configure for you

  • OLLAMA_HOST kept on the private interface; only the authenticated proxy is public
  • OLLAMA_KEEP_ALIVE tuned so your main model stays loaded between requests
  • OLLAMA_CONTEXT_LENGTH set to what the plan’s RAM can hold for your model
  • OLLAMA_NUM_PARALLEL and OLLAMA_MAX_LOADED_MODELS sized so models do not fight for memory
  • OLLAMA_ORIGINS set for any browser front-end on another domain
  • Your chosen models pulled at setup, for example llama3.1:8b plus an embedding model

Which plan fits

Starter (8 GB, 4 vCPU) holds one 7-8B Q4 model such as Llama 3.1 8B or Qwen 7B plus a small embedding model, at roughly 5-15 tokens/s. Plus (12 GB, 6 vCPU) fits a 13B Q4 model, or keeps an 8B chat model and a second model loaded together. Neither plan runs 30B-class models usefully.

The honest limits

These are CPU plans, and prompt processing is the slow part: a long RAG prompt can take many seconds before the first token appears. Realtime products and 70B models need GPU hardware, which we do not currently sell.

What people run Ollama for

Private chat + RAG

Pair with Open WebUI or AnythingLLM for a ChatGPT-style UI over your documents.

Automation backend

Give Activepieces or Node-RED a flat-cost local model for classification and extraction.

Dev & staging endpoint

A stable OpenAI-compatible endpoint for testing without burning paid tokens.

Compliance workloads

Keep prompts and outputs on hardware you control.

How Ollama hosting compares

OptionPriceWhat you get
Elestiofrom $30/moManaged Ollama; GPU only through a third-party add-on.
Cloud GPU instances$400–1,000/mo + egressThe right answer for realtime or 70B work — and priced accordingly.
AIHostBoxfrom $16.99/moDedicated single-tenant instance, managed updates and backups.

We are the cheap end of this table on purpose: quantized 7-13B models on dedicated CPU cores. If you need GPU speed, one of the others is the better buy and we will tell you so.

Prices verified September 2026 and may have changed since.

Licence and version policy

ItemPosition
Upstream licenceMIT
Sold as managed hostingYes
TenancySingle tenant — one customer per instance
PatchingCritical advisories within 24 hours, high severity within 72

Our full position on licences, including the applications we decline to sell, is on the licensing page.

Reading on Ollama

ComparisonOllama vs vLLMComparisonOllama vs LM StudioComparisonLiteLLM vs OpenRouterGuide · Private LLMRunning an LLM on CPU: what actually worksGuide · ArchitectureA self-hosted AI stack that actually holds togetherGuide · CostCutting an LLM bill without cutting qualityGuide · Team chatPrivate ChatGPT for your team, on a server you control

Works well with Ollama

Open WebUI logo

Open WebUI

A ChatGPT-style workspace for your team, pointed at any model you like.

from $5.99/moDetails →AnythingLLM logo

AnythingLLM

A private document workspace — drop files in, ask questions, get cited answers.

from $18.99/moDetails →OpenClaw logo

OpenClaw

Your always-on personal AI assistant, running 24/7 on a server you control.

from $9.99/moDetails →LiteLLM logo

LiteLLM

One OpenAI-compatible endpoint in front of every model provider you use.

from $5.99/moDetails →

More LLM runtimes we host

Hermes logo

Hermes

Nous Research’s Hermes models, served privately behind your own API.

from $16.99/moDetails →LocalAI logo

LocalAI

A drop-in OpenAI replacement for text, images, audio and embeddings.

from $16.99/moDetails →KoboldCpp logo

KoboldCpp

A single-binary GGUF server with its own UI and a KoboldAI API.

from $13.99/moDetails →llama.cpp logo

llama.cpp

The reference GGUF inference server, with nothing between it and the CPU.

from $13.99/moDetails →
Questions

Ollama hosting: your questions

Which models can I actually run?
On 8 GB: quantized 7-8B models such as Llama 3.1 8B, Hermes, Qwen 7B or Mistral 7B — the sweet spot for CPU. On 12 GB: 13B quantized. Larger models need GPU hardware, which these plans do not include.
How fast is CPU inference?
Roughly 5-15 tokens per second on 7-8B Q4 models on dedicated cores. Comfortable for chat, background automation and RAG; not for latency-critical products.
Can I use it with Open WebUI or SillyTavern?
Yes — point any Ollama-compatible front-end at your endpoint. We host both if you would like us to run the front-end as well.
Is the Ollama API protected?
Ollama itself has no authentication and binds to 127.0.0.1:11434 by default. We keep it on the private interface and publish it through a TLS reverse proxy that requires an access key, so only your apps reach the API. On your own servers, never expose OLLAMA_HOST=0.0.0.0 on an open port.
Why is the first reply after a pause slow?
Ollama unloads a model after OLLAMA_KEEP_ALIVE, five minutes by default, and the next request waits while the weights load from disk into RAM. We can raise the keep-alive, or set it to -1, so your main model stays resident, at the cost of that memory being held permanently.
Can I bring my own GGUF file or custom Modelfile?
Yes. Upload the GGUF over SFTP, write a Modelfile starting with FROM ./your-model.gguf, add a SYSTEM prompt or PARAMETER lines, and run ollama create. The result behaves like any library model through /api/chat and the OpenAI-compatible /v1 endpoints.
How much context can I use?
Ollama defaults to a 4096-token context (OLLAMA_CONTEXT_LENGTH). You can raise it, but the KV cache grows with context and competes with the weights for RAM. On the 8 GB Starter plan an 8B Q4 model with 8k context is comfortable and 16k is tight; long-document RAG works better with chunking than with a huge window.

Run Ollama without running a server

A dedicated Ollama instance, deployed and maintained by us. Month to month, cancel whenever.

Browse the catalogue Talk to a human