Open WebUI

Run Open WebUI Privately on high-performance GPUs

Open WebUI and GPU Trader

Open WebUI is an open-source, browser-based interface for interacting with large language models (LLMs) like those running on Ollama, LM Studio, or local GGUF models. Designed for speed, simplicity, and privacy, Open WebUI offers a clean chat interface with support for multi-user access, prompt history, system prompts, and customizable personas. It runs entirely on your own hardware or containerized environment, ensuring full control over data and model execution.

GPU Trader offers a pre-configured managed template to deploy Open WebUI in a single click—making it easy to run private, performant LLMs without needing cloud accounts or API keys.

Install Open WebUI on GPU Trader

1. Sign Up and Create an Account

   Start by signing in or creating an account on GPU Trader. If you need help, watch the GPU Trader Quickstart or read the docs.

2. Find and Rent a Compatible Instance

   Use the Find a Device page to browse available GPUs. Filter by model, price, or provider to select a system that meets Open WebUI’s requirements.

   Once selected, rent the instance directly from the dashboard.

   Remember that Open WebUI is an interface to models. The engine that allows you to use them is Ollama and the models have different infrastructure requirements. Review the Model Compatibility table below if you are unsure what instance to pick

   Need more help? Read the docs on finding and renting an instance.

3. Configure Your Instance for Open WebUI

   After renting your instance, configure it with GPU Trader’s pre-built template Open WebUI template. To deploy them template click “Add Stack”, browse and select the Open WebUI template. Click “Use Template.”

   Template:

   services:
     ollama:
       volumes:
         - ./volume1/docker/ollama:/root/.ollama:rw
       pull_policy: always
       tty: true
       restart: always
       image: ollama/ollama:latest
       ports:
         - 11434-11534:11434
       labels:
         - io.gputrader.ports.name.11434=Ollama API
       environment:
         - NVIDIA_VISIBLE_DEVICES=all
         - NVIDIA_DRIVER_CAPABILITIES=compute,utility
     open-webui:
       image: ghcr.io/open-webui/open-webui:cuda
       restart: always
       volumes:
         - ./volume1/docker/open-webui:/app/backend/data:rw
       depends_on:
         - ollama
       environment:
         - OLLAMA_BASE_URL=http://ollama:11434
         - WEBUI_SECRET_KEY=
       ports:
         - 8080-8180:8080
       labels:
         - io.gputrader.ports.name.8080=Open WebUI
       extra_hosts:
         - host.docker.internal:host-gateway
   

   Need more help? Read the docs on managing an instance or templates.

4. Launch Open WebUI and Start Conversing

   Once Open WebUI is installed, launch the application by selecting the port dropdown or kebab menu to access the UI from your browser. Now you can create your account, add new users, and download the models of your choosing.

Get Started with Open WebUI

We will be the first to tell you, we are not experts in Open WebUI or the models that run on it, even though we think it is really useful and fun. If you are new to LLMs it is best to refer to the documentation for that model. To help you understand model compatibility we have created a list. The list is not comprehensive and should be used as a reference. There is good information on the web dedicated to optimizing the performance of your LLM.

Model Compatibility

LLM	Best For	Recommended GPU(s)	Notes
DeepSeek R1	Logical reasoning, math, multilingual tasks	2x A100 80GB or 1x H100 80GB	Efficient Mixture-of-Experts model; rivals GPT-4 in performance.
Qwen 2.5-72B	Coding, multilingual tasks, long-context understanding	1x H100 80GB or 2x A100 80GB	Excels in coding and supports 128K context window.
Mistral 7B	Chatbots, summarization, lightweight applications	1x A10 24GB or better	Fast and efficient; ideal for real-time applications.
Gemma 2 27B	Multilingual QA, RAG, fine-tuning	1x A100 80GB or 2x A40	Optimized for performance on NVIDIA GPUs; good for RAG pipelines.
LLaMA 3.3 70B	General-purpose reasoning, coding, assistant tasks	2x A100 80GB or 1x H100 80GB	Strong all-rounder; extensive community support.