SELF-HOSTED · MULTIMODAL API

Image and text embeddings, on your infrastructure.

Build semantic search, image similarity and multimodal retrieval with two dedicated embedding models behind an authenticated API.

Embedding requests require a Bearer token. Public health and model metadata do not confirm GPU readiness.

ARCHITECTURE
Your clientHTTPS
CPU gatewayFastAPI · Auth
On-demand GPU workersModal
Qwen3-VL-Embedding-8BL40S
WeMM-Embedding-9BL40S
API OVERVIEW
GET /health   ·   GET /v1/models

Public metadata · Bearer required for inference

Two models. Separate vector spaces.

Choose the model and embedding settings that fit your application. Vectors from different models are not interchangeable.

MODEL 01 · QWEN

Qwen/Qwen3-VL-Embedding-8B

qwen3-vl-embedding-8b

Image + text embeddings

MODEL 02 · TENCENT

tencent/WeMM-Embedding-9B

wemm-embedding-9b

Image + text embeddings

Start building

Everything you need to deploy and connect your own client.

01

Deploy the service

Configure a Modal workspace, pinned checkpoints and private Secrets.

Deployment guide ↗