Alibaba Cloud: Qwen3-VL-Rerank

Qwen3 VL Rerank reorders retrieved candidates using visual and textual relevance. It is useful after an initial search, when a system needs to choose which mixed-media results best answer the user's query.

Compare models

qwen3-vl-rerank

AvailableAlibaba CloudRerankSync

Input / Output: $0.10 / —
Context: 120K
Modalities: VisionReranking

About Qwen3-VL-Rerank

Qwen3-VL-Rerank is Alibaba's multimodal reranker. Given a query that can be text or an image, it scores candidate documents that can contain text, images, or videos, and returns them ordered by relevance. It sits after a first-stage vector search to pick the best mixed-media results. Unlike Qwen3.7-Text-Rerank, which handles text only, it supports 33 languages and accepts image and video candidates.

Where it works well

Ranks documents that mix text, images, and videos in one request.
The query itself can be an image, enabling search by screenshot or photo.
Supports 33 major languages for queries and documents.
A custom instruction can describe what relevance means for your task.

When to choose another model

It covers 33 languages, while Qwen3.7-Text-Rerank covers far more for plain text.
Document counts are capped per request, with fewer images and videos than text passages.
Videos must be passed by URL, and not all image formats are accepted.

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.

RerankingTokenLab endpoint

POST/v1/rerank

curl -X POST "https://api.tokenlab.sh/v1/rerank" \
  -H "Authorization: Bearer sk-xxx" \
  -H "Content-Type: application/json" \
  -d '{
  "operation": "rerank",
  "model": "qwen3-vl-rerank",
  "query": "What is the capital of France?",
  "documents": [
    "Paris is the capital of France.",
    "Berlin is the capital of Germany."
  ]
}'

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 Oct 16, 2026, 12:08 PM 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-Rerank in Console with a prompt ready to edit or send.

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

Use cases

Best for

Vision

Product image search
Rerank product items against a customer's photo so the closest visual and textual matches come first.
Multimodal RAG
Order screenshots, slides, and passages retrieved for a question before they go to the answering model.
Video clip lookup
Pick the clips that best answer a text query from a pool of candidate videos.

Prompt examples

Query: red enamel dutch oven with a lid. Rank these product photos and descriptions by relevance.

Instruction: prefer documents that show the error dialog. Query: a screenshot of the failed upload.

Rank these five lecture clips for the question 'how does gradient clipping work'.

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 does Qwen3-VL-Rerank do?

It reorders candidate results by relevance to a query. Both query and candidates can include text, images, and video, so it works for mixed-media retrieval where a text-only reranker cannot judge the picture.

How is it different from Qwen3.7-Text-Rerank?

Qwen3-VL-Rerank accepts images and videos and covers 33 languages. Qwen3.7-Text-Rerank takes text only but handles far more languages and many more documents per request. Choose by whether your candidates are visual.

What media formats does it accept?

Images in common formats such as JPEG, PNG, and WEBP, passed by URL or Base64, and videos in MP4, AVI, or MOV, passed by URL only.

Can I steer what counts as relevant?

Yes. An instruction parameter lets you say what relevance means, for example matching a specific error or style. It is optional, and results return as indexes with relevance scores.

Where in a pipeline does it belong?

After retrieval. Use embeddings or keyword search to gather candidates, then rerank the shortlist and pass the top results on. It is not a search index by itself.

How much does Qwen3-VL-Rerank 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-Rerank use?

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

Which operations does Qwen3-VL-Rerank support?

Qwen3-VL-Rerank supports Reranking. Select an operation above to see its endpoint and request example.