Managing Models
View, rename, organize, and understand your fine-tuned models.
The models panel
Your dashboard's left panel shows all your fine-tuned models. Each card displays:
- Display name — editable label (defaults to the base model name)
- Description — your original use-case description
- Provider badge — OpenAI, Gemini, or GPU
- Base model — which LLM was fine-tuned
- Status badge — current state (Queued, In Progress, Succeeded, Failed)
- Timestamp — when training completed (for finished models)
Actions
Chat
Click Open in chat on any succeeded model to start a conversation.
Rename
Give your models descriptive names for easy identification. Click the rename action on the model card.
Good naming conventions:
- Include the purpose: "Support Agent - Help Docs v2"
- Include the audience: "Blog Writer - Formal Tone"
- Include a version: "Code Review Bot v3"
Download adapter
For Qwen fine-tunes, download the LoRA adapter as a .zip file. See LoRA Adapters.
Refresh
Click the refresh button to force-check for status updates. The dashboard also polls automatically.
Model limits
| Plan | Regular models | Premium models |
|---|---|---|
| Free | 3 | — |
| Pro | 15 | 5 |
| Enterprise | 100 | Custom |
Premium models are base models that require a paid plan — currently only Gemini 2.5 Pro.
When you reach your model limit, you'll need to upgrade or wait for existing models to be removed before creating new ones.
Understanding model status
| Status | What it means | What to do |
|---|---|---|
| Queued | Job is waiting to start | Wait — it'll start shortly |
| Validating files | Data is being processed | Wait — this takes 1–5 minutes |
| In progress | Training is running | Wait — 5–45 min depending on model |
| Succeeded | Ready to use | Open in chat or call via API |
| Failed | Training didn't complete | Check your data and try again |
Troubleshooting failed models
Not enough data Your files didn't contain enough usable text to train on. Try adding more data or a different file format.
Provider error A transient issue with OpenAI, Google, or the GPU cluster. Create a new job with the same data — it usually works on retry.
Invalid data format The cleaned data didn't pass the provider's validation. Try a different file format (e.g., convert PDF to plain text) or simplify your data structure.