Create a Dedicated Deployment

Prerequisites

Before deploying, make sure you have:

1. A Mycelis account — register at mycelis.ai if you don't have one yet
2. A workspace — create one after login
3. Credit in your wallet — navigate to Dashboard → Wallet to top up

Dedicated GPU deployments are billed hourly from the moment the deployment is active, regardless of actual usage. Stop or remove the deployment when you no longer need it to avoid unnecessary costs.

Navigate to the Model Marketplace

Click Models Marketplace in the left sidebar. You'll find hundreds of current models. You can:

• Search by name using the search bar at the top
• Filter by category — text generation, vision, embeddings, code, and more
• Load any vLLM-compatible Hugging Face model if the model you need isn't listed yet

If a model you need is missing, let us know at mycelis.ai/feedback or info@mycelis.ai.

Select your model and click Add Model to open the deployment wizard.

The Deployment Wizard

The wizard walks you through four steps.

Step 1 — Scaling & Schedule

Configure how the model scales and when it runs.

User scaling: Based on your concurrent request requirements, Mycelis automatically selects the most suitable GPU for you.

Daily schedule: Define a time window for the deployment to be active (e.g., 08:00–22:00). Leave it empty for continuous operation — the deployment runs until you stop or remove it.

Step 2 — Optional: LoRA Fine-Tuning

Skip this step if you don't need fine-tuning. To fine-tune the model before deployment, provide a link to a JSONL dataset:

• Mycelis starts a LoRA training job before the deployment
• A more powerful GPU is used temporarily for training
• After training, the fine-tuned model is saved under My Models
• The deployment uses your customized model variant

Step 3 — Optional: OpenWebUI

Mycelis can host a private OpenWebUI instance for you and your team. Link your deployment here to make the model accessible in a chat interface.

Notes:

• OpenWebUI must first be activated under Workspace Settings → OpenWebUI
• You can add or remove the link at any time, even after deployment creation
• The OpenWebUI instance is only active when at least one deployment or agent is linked to it

Step 4 — Overview & Create

Review your full configuration — model, GPU type, schedule, and estimated costs. Click Create Deployment.

Depending on model size, your deployment will be ready in 1–3 minutes. Once active, it appears in your workspace with a live API endpoint.

Next steps

• Configure an Agent — add routing, Knowledge Bases, and MCP Tools on top of your deployment
• Create a commercial deployment (Managed Keys) — use OpenAI, Anthropic, or Google models
• Create a commercial deployment (BYOK) — use your own provider key