Side by side
| Fine-tuning it yourself | Facet | |
|---|---|---|
| What you bring | Training examples, often thousands, written or labeled by you | A description and a few real inputs |
| Checking quality | Build your own test set and scoring | Facet drafts the rules and examples, then checks every version |
| Compute | Rent GPUs or a hosted service, and pick the settings | Nothing to manage; each build is quoted up front |
| Hosting | Deploy and scale it yourself | Callable with the OpenAI SDK as soon as it’s built |
| Improving it | Gather more data, train again, compare by hand | Say what’s wrong; Facet quotes, builds and checks the fix |
| First result | Setup, data and a training run come first | Usually about half an hour after you describe the job |
| Control | Every setting is yours | Export the model any time, free, and run it yourself |
When fine-tuning it yourself is the better choice
- You need a specific training method or model that Facet doesn’t offer
- The work involves images, audio or video; Facet handles text today
- Your team wants to control every training setting
- Data can never leave your own hardware, not even to build the model
Try Facet with your agent
Connect Claude, Cursor, ChatGPT, Codex or VS Code to Facet, then paste:
Use Facet to build a specialist for this job: ... Here are a few real inputs: ... Show me
the draft and the price before you approve anything.
Questions
Is Facet just fine-tuning?
Fine-tuning is one tool Facet uses. Facet also writes the examples, checks every version against your rules and hosts the result, and when you ask for a change it first tries improving the instructions, which costs cents, before it trains again.
Which models does Facet start from?
Facet starts from open-weight Qwen3.5 models in two sizes and picks the one that fits your task. You can change the pick before it builds.
Can I take my model somewhere else?
Yes. Export it any time, free. You get a LoRA adapter for its Qwen3.5 base model, a standard open format that tools such as vLLM run, along with its instructions.