One key, many models: how to read aijiau prices without a scoreboard
High cost-performance is positioning, not an unpublished speed test. Keep Jev customer price, TypeSafe list price, and chat catalog prices apart. Points to the console and /jev.
Agents often need several chat models plus a decision layer. aijiau’s shape is one sk-... key and a catalog on api.aijiau.com. The www explainer is /gateway. The public English name is LLM API Gateway / Unified AI API.
Keep three numbers apart
| Figure | What it is | Source |
|---|---|---|
| ¥0.6 / 1M input, output free | aijiau customer price for Jev / System One on the live API host | published aijiau customer price |
| $0.042 / 1M input, output free | TypeSafe vendor list price | TypeSafe docs and product pages |
| Chat model unit prices | whatever the console lists that day | api.aijiau.com |
Do not mash these into a “we are N× faster” table. This site does not publish a self-measured QPS, and it does not publish a Chinese accuracy percentage.
What one key is for
- Fewer upstream accounts and fewer secret copies.
- OpenAI-compatible
/v1/chat/completionsfor models that write. - The same key on
POST /v1/systemonewithmodeljevfor routing and gates. - Login, Playground, and billing stay on the API host. www is not a second console.
The chat catalog moves. Trust the console, not this note, for model ids.
How this sits next to /jev
/jev is the System One flagship: Choice / Score / Noul, pasteable samples, and the Chinese / Traditional self-test. This post is only how to use the gateway and how to read prices. Do not dump tax quoting into this channel.
- Learn the decision layer → /jev
- Land the first System One call → console · walkthrough
- Machine shape → /jev/docs
- Chat getting started → OpenAI-compatible note
CS: [email protected].
