通义千问 embedding 系列 4B 档——精度与成本的均衡位，RAG 知识库与语义检索的常用默认选。
| No. | M36 |
| Vendor | Alibaba |
| Type | 嵌入 |
| Parameters | 4B |
| Billing | Pay-as-you-go · per token |
| Official reference price | Open-source free · same as HF |
| AIXX Cloud price | Free · direct cloud calls |
| Use case | Semantic search / RAG knowledge bases / similarity ranking |
| Official API direct | Self-hosted relay | AIXX Cloud | |
|---|---|---|---|
| Onboarding cost | Foreign card + VPN signup | Server + ops + renewal anxiety | One sentence to your agent |
| Switching models | One account per vendor | Roll your own routing | One Key, every model |
| Failover | Single point of failure | Write it yourself | Multi-channel auto-routing |
| Interface | Every vendor, its own flavor | Smooth it yourself | Unified OpenAI-compatible |
No relay setup — tell your agent "run npx aixxai" to install AIXX, then just assign tasks and name Qwen3-Embedding-4B; the OpenAI-compatible interface routes automatically.
Free · direct cloud calls (Open-source free · same as HF). Pay-as-you-go, no monthly fee; usage details are visible in the console anytime.
No foreign credit card, no VPN signup, one Key for every model; pricing follows the live rates in the console.
Fully supported on the OpenAI-compatible interface (stream / function calling / JSON mode) — plug-and-play with the agent ecosystem.
Tell your agent:
npx aixxai install ┄┄┄┄┄┄