← Managed inference

Qwen3-VL Embeddings (4096 dim, multimodal)

Puts text, images, screenshots and video in one vector space

$1 in free credits

Overview

Qwen3-VL-Embedding-8B is built on Qwen3-VL-8B-Instruct for multimodal retrieval. It embeds text, images, screenshots, video and arbitrary mixtures of them into a shared space and covers 30+ languages. Knowledge bases here store its full 4096-dimension vectors. Reach for it when a knowledge base has to answer an image query with text documents, or the reverse.

Strengths

  • One shared space for text and images
  • Full 4096-dimension vectors
  • Instruction-aware embeddings

Use cases

  • Knowledge bases mixing images and text
  • Screenshot search
  • Cross-modal retrieval such as a text query matching an image

Good to know

Used through knowledge bases or the workspace embed endpoint, which takes text only, so images reach it through a knowledge base; the model API has no embeddings route. The output size is fixed at 4096. At 8B parameters it is much heavier than bge-m3, so choose it only when images or video genuinely need to be in the index.

API access

One OpenAI-compatible API for every model in the catalog, plus the CLI and the function SDK inside workflows. Sign up and try $1 of inference free, then get $10 more once you've spent $5, and these requests work:

qwen3-vl-embedding · curl
# qwen3-vl-embedding is an embedding model; the model API has no embeddings route.
# A knowledge base uses it when you pick it at creation.
curl https://model.graphn.ai/v1/models \
  -H "Authorization: Bearer $GRAPHN_API_KEY"