Qwen3-VL-Embedding-2B

RK1820   |   LlamaPi Ready

Model Overview

Run with LlamaPi

llamapi pull qwen3-vl-embedding:2b
llamapi load qwen3-vl-embedding:2b

Use the loaded model through LlamaPi's embeddings API. See the LlamaPi Wiki for setup and usage.

About the Original Model

Qwen3-VL-Embedding-2B is part of the Qwen3-VL-Embedding series built on the Qwen3-VL foundation model, designed for multimodal information retrieval and cross-modal understanding. It generates high-dimensional vectors from text, images, screenshots, videos, and mixed-modality inputs for retrieval and clustering.

• Accepts text, images, screenshots, video, and arbitrary multimodal combinations in a unified framework, delivering state-of-the-art multimodal retrieval performance (73.2 overall on MMEB-V2) at only 2B parameters.
• Generates semantically rich vectors in a shared visual-textual space, with a 32K context window and embedding dimensions up to 2048 (user-definable from 64 to 2048 via MRL).
• Supports 30+ languages and is instruction-aware, with customizable task instructions and strong performance even with quantized embeddings.

License & Attribution

The models provided in this repository are converted and adapted by the Firefly AI team from the original models. All intellectual property rights and copyright in the underlying original models remain with the original official team, and their distribution is governed by the corresponding official open-source license.