AI and agents · Rio de Janeiro or São Paulo

VPS with Ollama and Open WebUI

Run your own AI on your VPS: Ollama runs the Qwen 3.5 model and Open WebUI gives you a browser chat with login and HTTPS. No API key, no per-message fees, and your conversations stay on your server.

Local AI, no third-party API
Qwen 3.5 already downloaded
Chat with HTTPS and login
Rio de Janeiro or São Paulo, Brazil

ollama.yourdomain.com

Open WebUI admin panel with the local model installed

Qwen 3.5 summarizing a customer's request in bullet points, running2 screens

Plans

VPS plans for running Ollama and Open WebUI

Language models use a lot of memory and CPU. The setup with Qwen 3.5 (4B) needs at least 5120 MB of free memory. We recommend the Plus plan to get started, and Pro gives you more headroom.

Location

Selected

Pro

US$ 21.95/mo

4 vCore CPU
6 GB Memory
NVME 120 GB Disk
Unlimited Traffic

Ask on WhatsApp

Recommended

Turbo

US$ 35.95/mo

4 vCore CPU
12 GB Memory
NVME 160 GB Disk
Unlimited Traffic

Ask on WhatsApp

Max

US$ 53.95/mo

6 vCore CPU
18 GB Memory
NVME 180 GB Disk
Unlimited Traffic

Ask on WhatsApp

Ultra

US$ 89.95/mo

8 vCore CPU
24 GB Memory
NVME 320 GB Disk
Unlimited Traffic

Ask on WhatsApp

Billed monthly in US dollars, with the app installed. Rio de Janeiro or São Paulo, same price. Plans start at the smallest size where the app runs well. Need more? Upgrade from the client area, without reinstalling.

Where the server lives

Ollama e Open WebUI in Rio de Janeiro

With Rio de Janeiro selected, your VPS with Ollama e Open WebUI runs in Mega's original data center, on our own network with a presence at the Rio internet exchange (IX.br) and redundant transit. It is the lowest latency for users in Rio, Espírito Santo and Minas Gerais, and our support team works in the same city as the server.

Measured latency, not a promisemeasured 19 minutes ago

Rio de Janeiro to

Vivo11.4ms

Claro19.5ms

TIM20.2ms

Oi11.1ms

Algar4.5ms

São Paulo to

Vivo15.0ms

Claro8.6ms

TIM23.7ms

Oi15.2ms

Algar7.0ms

Ping from real customer connections on each carrier to our data centers in Rio and São Paulo, measured every hour by independent probes (Globalping). Algar: measured from our edge, where we connect to it directly.

What Ollama is and what Open WebUI does

Ollama is open-source software that runs language models (LLMs) on your own server. It downloads the model, keeps it in memory and answers through an API, without relying on OpenAI, Google or any other external service.

Open WebUI is the chat interface that sits in front of Ollama. It stores conversation history, supports multiple users, lets you upload documents for the AI to read and switches models in a few clicks.

This VPS has no GPU, so all processing runs on the CPU. That is why the model is a small one, Qwen 3.5 with 4 billion parameters. It handles summarizing, rewriting, classifying and drafting well, but it gets facts wrong more often than the large cloud models.

Source: Official Ollama GitHub repository, accessed October 2026

What comes installed on your VPS

Debian 13 with Docker
Ollama 0.40 in a container that restarts on its own after a server reboot
Qwen 3.5 (4B) model downloaded and loaded in memory
Open WebUI with an administrator account created and public sign-up turned off
Your own HTTPS address with a Let's Encrypt certificate
Root access via SSH and the VPS panel in the client area
Access details sent by email and WhatsApp

Ollama use cases on your own server

Summaries of long texts

Paste an email, meeting minutes or a report and ask for the key points as a list.

Draft replies

Generate suggested replies to customers and adjust the tone as needed.

Confidential information

Work with contracts and internal documents without sending them to third parties.

Automations with no per-call cost

Tools like n8n can query the Ollama API as often as you need.

Triage and data extraction

Sort requests by topic or pull specific fields out of a text.

Compare open models

Download other models from the Ollama library and see which one works best for your case.

Ollama and Open WebUI in action

Qwen 3.5 listing a customer's request as bullet points, on a Mega test VPS (example).

Open WebUI admin area with the local model installed (example).

Screenshots taken on a Mega Dedicados test installation, with fictional data and no customer information.

What comes already configured

Ollama 0.40

Pinned version, with the model already loaded in memory so replies don't wait on a model load.

Qwen 3.5 (4B)

Open model that handles English and Portuguese well, downloaded during installation.

Open WebUI

Browser chat with login and HTTPS on your own address.

Direct answers

Reasoning mode is turned off so the CPU can reply faster.

No background tasks

Automatic titles and suggestions are disabled so they don't compete with the chat for CPU.

Ollama API closed

Ollama is not exposed to the internet; only Open WebUI talks to it.

From order to first use

1
Choose your plan and location
Select Plus or Pro with the Ollama and Open WebUI app and pick whether the VPS is in Rio de Janeiro or São Paulo, Brazil.
2
Pay
Pay with Pix, card or boleto, on a monthly, semiannual or annual cycle. Installation starts as soon as payment is confirmed.
3
Wait for the installation
The system creates the VPS, installs Ollama and Open WebUI and downloads the model, about 13 GB in total. In our tests it took 5 to 15 minutes.
4
Open the chat
Address, login and password arrive by email and WhatsApp. Just open the address in your browser and start chatting.

Which Mega plan to use with Ollama

Start with Turbo, the recommended plan, which gives you headroom over the 5120 MB of free memory required. Pro works for testing, and Max or Ultra suit more users or larger models.

Plan	Best for
Pro 4 vCore · 6 GB	Good for testing Qwen 3.5 (4B) on your own with a few conversations, running close to the required memory limit.
Turbo 4 vCore · 12 GB	The recommended plan: double the memory of Pro, leaves room for the system and handles daily personal use well.
Max 6 vCore · 18 GB	For those sharing the chat with family or a small team who want more CPU and memory headroom.
Ultra 8 vCore · 24 GB	For heavy use with several people at once, or for testing larger models with plenty of spare resources.

Frequently asked questions

Frequently asked questions about Ollama and Open WebUI

How much does Ollama and Open WebUI hosting cost?

A VPS with both apps installed starts at per month on the Plus plan, and Pro costs US$ 21.95. There are no per-message fees and no API key is needed.

Is self-hosted Ollama as good as ChatGPT?

No. A model running on a server without a GPU is much smaller than the cloud ones. It works well for summarizing, drafting and classifying text, but it can make up information, so check anything critical.

How fast is a VPS with Ollama and Open WebUI?

In a test with 4 vCPUs and 6 GB, text started appearing in about 1 second and came out at around 7 tokens per second, slightly above reading speed. A paragraph usually takes under a minute.

Can I install other models on my Ollama VPS?

Yes. Download them from the Open WebUI admin panel or over SSH with docker exec ollama ollama pull followed by the model name. Larger models need more memory and answer more slowly.

Does self-hosted Ollama and Open WebUI send my chats to an AI company?

No. The model runs inside your VPS and no message is sent to OpenAI, Google or any other service. Your data stays in the data center you chose.

Can I turn on reasoning mode?

Yes, in the model's advanced parameters. It is off by default because, without a GPU, reasoning makes each answer much slower.

Who updates Ollama on the VPS?

The installed version is pinned so that no update changes behavior without warning. You update whenever you want, through root access, after testing.

How do I start over if something goes wrong?

Use the reinstall option in the VPS panel, in the client area, which erases the server's data. If you want the app installed again, 24/7 support in Portuguese can help by WhatsApp, ticket or phone.

Updated on October 19, 2026