Alibaba Cloud: Qwen3-VL-Embedding

Qwen3 VL Embedding maps visual and textual content into vectors for retrieval. It is useful for finding images with a text query or connecting a visual asset with related written information.

Compare models

qwen3-vl-embedding

AvailableAlibaba CloudEmbeddingSync

Input / Output: $0.10 / —
Context: 32K
Modalities: EmbeddingVision

About Qwen3-VL-Embedding

Qwen3-VL-Embedding is Alibaba's multimodal embedding model: it maps text, images, and video into vectors so one query can retrieve any of them. Alibaba lists 33 major languages and selectable output sizes from 256 to 2560 dimensions, plus a fusion mode that merges several inputs into one vector. It is the more flexible of Alibaba's two multimodal embedding models.

Where it works well

Embeds text, images, video, and several images in one model, enabling cross-modal search.
Selectable output sizes from 256 up to 2560 dimensions let you trade accuracy against storage.
Fusion mode combines all inputs of a request into a single vector, such as an image plus its caption.
Covers 33 major languages according to Alibaba's reference.

When to choose another model

Large 2560-dimension vectors increase index size, so pick a smaller size if storage matters.
It produces vectors only; ranking results or generating answers needs another model.

Getting started

Create API key
Create a key in Console, then use it with every model on the platform.

Send your first request

Copy the example for your language and run it against the endpoint.

Multimodal embeddingTokenLab endpoint

POST/v1/multimodal/embeddings

curl -X POST "https://api.tokenlab.sh/v1/multimodal/embeddings" \
  -H "Authorization: Bearer sk-xxx" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "qwen3-vl-embedding",
  "input": {
    "text": "The quick brown fox jumps over the lazy dog."
  }
}'

Parameters

Multimodal embedding

Parameter	Values and limits	Default	Notes
`dimensions`	Not provided	Not provided	This field is accepted. The current public contract does not specify a complete value range or list. No additional values are promised. The omitted-field default is not verified.
`input`	Not provided	Not provided	Not provided
`input_modality`	Not provided	Not provided	Not provided

TokenLab currently keeps image input behind ENABLE_MULTIMODAL_IMAGE_INPUT; public contract should not promise image/video input until live probes and docs are opened end to end.

Pricing

The Verified price applies to Verified, which costs less on most models. The Official price is the model maker's published price and applies to the more reliable Official route. Auto bills the route that completes the request.

Default

per 1M tokens

Official price: Input $0.10 / Text Input $0.10 / Image Input $0.258
Verified price: —
Discount: —

	Official priceper 1M tokens	Verified priceper 1M tokens	Discount
Default	Input $0.10 / Text Input $0.10 / Image Input $0.258	—	—

Usage & activity

Success rate is the share of requests that completed. Latency is how long a full response takes; P95 means 95% of requests finished within that time.

Usage & availability

Last 24 hours

Requests
Success rate
P95 latency
Total tokens

Model performance

Last updated Nov 1, 2026, 8:13 AM UTC

Metrics appear once privacy and data volume thresholds are met.

Data is based on aggregate user requests, excluding status checks.

Open in Console

Open Qwen3-VL-Embedding in Console with a prompt ready to edit or send.

Help me try qwen3-vl-embedding with multimodal-embedding at /v1/multimodal/embeddings. Show the result and cost.

Use cases

Best for

Vision

Image search by text
Index a photo library as vectors and retrieve pictures from a written description in several languages.
Product catalogues
Fuse each product photo with its title into one vector so shoppers find items by look or by words.
Video retrieval
Embed video clips and let editors find footage with a plain-text query, instead of tagging every clip by hand.

Prompt examples

Embed this product photo together with its title text as a single fused vector.

Embed the query 'red leather sofa in a bright living room' for image retrieval.

Embed this 20-second video clip so it can be found with text searches.

This model has conditional pricing. Monthly token totals alone cannot produce a reliable estimate; use the detailed pricing for the request specification, cache, and applicable time window.

FAQ

What can Qwen3-VL-Embedding embed?

Text, images, video, and multiple images. Alibaba's reference lists both independent embeddings per input and a fused mode that returns one vector for all inputs together.

What output dimensions are available?

2560 by default, with 2048, 1536, 1024, 768, 512, and 256 as alternatives. Smaller sizes save index space at some cost in retrieval precision. Test a smaller size on your own queries first, since it often retrieves nearly as well.

Which languages does it support?

Alibaba lists 33 major languages, including Chinese, English, Japanese, Korean, French, and German, so queries in many languages can find the same images. Check niche languages on samples before launch.

How does it differ from Tongyi Embedding Vision Plus?

Qwen3-VL-Embedding offers adjustable dimensions, a fused mode, and broader language coverage. Tongyi Embedding Vision Plus has one fixed output size and is limited to Chinese and English.

How much does Qwen3-VL-Embedding cost?

Rates depend on the billing unit, specification, and usage. Compare matching conditions in the model's detailed pricing; a single rate does not determine the total cost.

Which endpoint should Qwen3-VL-Embedding use?

Use https://api.tokenlab.sh/v1/multimodal/embeddings for Qwen3-VL-Embedding. The request example below shows the matching code shape.

Which operations does Qwen3-VL-Embedding support?

Qwen3-VL-Embedding supports Multimodal embedding. Select an operation above to see its endpoint and request example.