SelfHostMap

Fully offline private AI chat (Ollama + Open WebUI)

Quick answer

Ollama downloads and runs open-weight models on your own hardware and exposes an API. Open WebUI connects to it and gives staff a chat interface with accounts and document upload. No prompt leaves your server, and there is no per-token bill. The trade-off is hardware: memory needs depend entirely on the model you load. None of Ollama + Open WebUI publishes an official CPU/RAM figure, so we do not estimate a server size or cost. Size the server for the models and workloads you plan to run.

Last updated · sizing from each app's official docs · prices from Hetzner Cloud (2026-10-22)

Components

  • Ollama: Runs open-weight models locally
    Official hardware: Not published (measured by us: 1.9 GiB RAM while working) · License: MIT (Permissive)
  • Open WebUI: Chat interface with accounts and document upload
    Official hardware: Not published (measured by us: 909 MiB RAM while working) · License: Open WebUI License (BSD-3-Clause style with branding clause) (Custom / conditional)

Combined hardware and monthly cost

Required components, official minimumsNone of the required components publishes an official CPU/RAM figure, so we do not estimate a server size.
Smallest server that fitsnot estimated
Measured RAM in our own testOllama 0.40.8: 1.9 GB peak while working; Open WebUI 0.12.3: 909 MB peak while working. Measured on SelfHostMap test VM (Oct 2026), 2026-10-22; container RAM only, not the whole server; no load test. Model-runner figures are for the small test models named on the app page; larger models need far more RAM.
Server sized from our measurementCPX32 (4 vCPU, 8 GB RAM, 160 GB disk) €35.99/mo Measured in our own test (2026-10-22): Ollama 0.40.8 1982 MiB + Open WebUI 0.12.3 909 MiB working RAM, x1.5 plus 512 MiB for OS and Docker = 4849 MiB, so CPX32 (8 GB). Light use, no load test. Not an official figure.

How we got this: we added up the official minimum vCPU and RAM each project publishes (linked on its app page) and took the cheapest Hetzner Cloud Regular Performance (shared AMD vCPU) plan that meets the total. Prices read from www.hetzner.com/cloud/regular-performance on 2026-10-22. "From" monthly price cap shown by Hetzner, excluding VAT. Prices vary by location. Covers the server only; it excludes LLM API usage, backups, extra storage, GPUs and your time. Running everything on one server is our simplifying assumption. Measured figures, where we have them, are on each app page.

Cost vs ChatGPT / Claude

1. The SaaS you pay for

2. Self-hosted stack

3. Model API spend (self-hosted side)
SaaS per year—
Self-host per year—
Break-even—

SaaS (cumulative)Self-host incl. setup (cumulative)
Open in full calculator
Assumptions and sources

    Deployment outline

    1. Run Ollama (Docker or the Linux install script) and pull a model. source
    2. For NVIDIA GPUs, check compute capability (5.0+) and driver version (550+; 570+ for older cards). source
    3. Run Open WebUI with --add-host=host.docker.internal:host-gateway so the container can reach Ollama on the host. source

    This is a summary of the official docs, not a full tutorial. Each app page has the exact official install command.

    Gotchas

    • Ollama: Older NVIDIA cards (compute capability 5.0–6.2) need driver 570 or newer. source
    • Ollama: With the Vulkan backend, Ollama needs root or the cap_perfmon capability (sudo setcap cap_perfmon+ep /usr/local/bin/ollama) to read free VRAM; otherwise it schedules models using approximate sizes. source
    • Open WebUI: Branding may only be changed for deployments of 50 or fewer end users in a rolling 30-day window, unless you have permission or an enterprise license. source
    • Open WebUI: If you only use hosted APIs, the :slim image is about 197 MB instead of 1.62 GB. It cannot be combined with the CUDA or bundled-Ollama variants. source

    What this does not replace

    • None of these apps includes an AI model. You connect a model API (billed by its provider) or run open models yourself, for example with Ollama.
    • Answer quality depends on the model you connect, not on the chat app.
    • You are responsible for the server, updates, backups and user management.

    FAQ

    How much RAM do I need to run AI fully offline?

    It depends on the model, not on Ollama. Neither Ollama nor Open WebUI publishes a fixed figure, so we do not give a server size here. Check the size of the model you plan to load.

    How much does a fully offline private ai (no cloud api) stack cost per month?

    None of Ollama + Open WebUI publishes an official CPU/RAM figure, so we do not estimate a server size or cost. Size the server for the models and workloads you plan to run.

    What hardware does Ollama + Open WebUI need?

    None of the required components publishes an official CPU/RAM figure, so we do not estimate a server size..

    Can a business use these apps commercially?

    Ollama: MIT (internal business use allowed). Open WebUI: Open WebUI License (BSD-3-Clause style with branding clause) (internal business use allowed). Summaries only, not legal advice.

    Is it cheaper than ChatGPT or Claude?

    It depends on team size, API usage and how you value your time. Use the calculator on this page; it starts with 10 users and ChatGPT Business (Standard seat) at the price published on 2026-10-22.

    Other stacks

    Not sure which app fits? Try the 5-question wizard.

    Every figure comes from the linked app pages and official docs. Spotted something wrong? Tell us.