Guide
·
10 min Lesezeit
Deployment Guide
Deployment types
| Type | Description | Cost |
|---|---|---|
| GPU Instance | Dedicated model on RunPod GPU | Hourly |
| Managed Keys | OpenAI / Anthropic / Google via Mycelis | Per token |
| BYOK | Your own provider key | Per token |
| Shared Pool | Shared infra for selected models | Per token |
1) GPU instance
Best for data sovereignty and predictable performance.
- Go to Dashboard → My Models → New Deployment
- Select GPU Instance
- Pick a base model (or trained model)
- Select GPU type and runtime schedule
2) Managed keys
Use OpenAI, Anthropic, or Google without managing external credentials.
- Select Managed Key
- Choose provider and model
- Deploy instantly (no provisioning delay)
3) BYOK
Use your own provider key through Mycelis.
- Select BYOK
- Choose provider (
openai,anthropic,google, or custom URL) - Store your API key securely