Side by side
| Frontier model | Specialist from Facet | |
|---|---|---|
| Getting started | Write a prompt and keep it working | Describe the job; Facet writes the instructions and examples |
| Quality on your task | Depends on the prompt; you build your own tests to know | Checked against your rules before every version goes live |
| Updates | The provider retires and replaces models on its own schedule | A new version goes live only after it passes your rules; one step rolls it back |
| Speed | A large general model answers every call | Built for one job, so it answers faster |
| Cost | Frontier prices on every call | A small specialist answers every call; each build is quoted up front |
| Improving it | Edit the prompt, then retest everything yourself | Say what’s wrong; Facet quotes the fix and checks it against every rule |
| Ownership | The model stays with the provider | Export it any time, free, and run it yourself |
When a frontier model is the better choice
- Open-ended work: writing, research, conversation
- Work that changes shape from one request to the next
- Low volume, where a good prompt is cheaper than building anything
- Jobs that need broad world knowledge or long chains of reasoning
- While you are still working out what the job is
Switching takes two lines
Keep the OpenAI SDK you use today. Point it at Facet and call your specialist by name:
import os
from openai import OpenAI
client = OpenAI(base_url=os.environ["FACET_BASE_URL"], api_key=os.environ["FACET_API_KEY"])
reply = client.chat.completions.create(model="support-triage", messages=[{"role": "user", "content": ticket}])Or ask your agent: connect Claude, Cursor, ChatGPT, Codex or VS Code to Facet and say “use Facet to build a specialist for this job, and show me the price first.”
Questions
Will a specialist be as accurate as a frontier model on my task?
Facet measures that before you rely on it. Every version of a Facet specialist is measured against your own rules and examples, you see how it does on each, and it goes live only when it meets the bar you set for every rule. A specialist is built to beat a general model on one narrow job with clear rules; on open-ended work it won't.
Can I use Facet and a frontier model together?
Yes. Keep a frontier model for open-ended work and move one high-volume job to a specialist. Both use the same OpenAI SDK, so your code calls each the same way.
How long does it take to switch a job to Facet?
Describing the job takes a few minutes. Facet builds the first version on its own, usually in about half an hour, and then you change two lines in your code.