Qwen3-VL-Reranker-2B
Qwen/Qwen3-VL-Reranker-2B
A popular open reranker model, with 300.3K downloads a month. gigarouter benchmarks and hosts it as an OpenAI-compatible API.
about this model
The Qwen3-VL-Reranker-2B is a specialized multimodal reranking model built on the Qwen3-VL foundation. It accepts a (query, document) pair where each may contain text, images, screenshots, videos, or arbitrary combinations of these modalities, and outputs a precise relevance score. Designed to work alongside the Qwen3-VL-Embedding model in a two-stage retrieval pipeline, it refines initial recall results to significantly boost retrieval accuracy.
Key Strengths
- Multimodal versatility: Handles text, images, screenshots, and videos within a unified framework, supporting image-text retrieval, video-text matching, visual question answering, and multimodal content clustering.
- High-precision reranking: Delivers superior relevance scoring compared to baseline rerankers and base embedding models across diverse tasks.
- Multilingual support: Covers over 30 languages, inheriting Qwen3-VL’s multilingual capabilities.
- Instruction-aware: Customizable input instructions for different tasks; evaluations show 1–5% improvement when using task-specific prompts (English recommended).
- Context length: 32K tokens.
Ideal Use Cases
Performing high-accuracy reranking in multimodal search pipelines, visual document retrieval, and cross-modal understanding applications where initial embeddings require refinement.
Benchmark Performance
| Model | Size | MMEB-v2 Retrieval (Avg) | MMEB-v2 Image | MMEB-v2 Video | MMEB-v2 VisDoc | MMTEB Retrieval | JinaVDR | ViDoRe v3 |
|---|---|---|---|---|---|---|---|---|
| Qwen3-VL-Embedding-2B | 2B | 73.4 | 74.8 | 53.6 | 79.2 | 68.1 | 71.0 | 52.9 |
| jina-reranker-m0 | 2B | – | 68.2 | – | 85.2 | – | 82.2 | 57.8 |
| Qwen3-VL-Reranker-2B | 2B | 75.1 | 73.8 | 52.1 | 83.4 | 70.0 | 80.9 | 60.8 |
| Qwen3-VL-Reranker-8B | 8B | 79.2 | 80.7 | 55.8 | 86.3 | 74.9 | 83.6 | 66.7 |
For further details, including full benchmark evaluation, hardware requirements, and inference performance, refer to the technical report, blog, and GitHub repository.
We're benchmarking and onboarding Qwen3-VL-Reranker-2B as a hosted, OpenAI-compatible API. Sign in for free credit and be ready when it lands, or tell us you want it and we'll prioritize it.