Jina AI 发布多模态嵌入模型 v5-omni-small
原标题:jinaai/jina-embeddings-v5-omni-small
AI 摘要
Jina AI 发布了 jina-embeddings-v5-omni-small,一个约 1.74B 参数的多模态嵌入模型,支持文本、图像、视频和音频输入,生成 1024 维共享向量空间嵌入,并与文本版 v5 模型对齐,无需重新索引即可跨模态检索。该模型支持检索、分类、聚类和文本匹配等任务,最大序列长度 32768,提供 Matryoshka 维度裁剪,并可通过 Elastic Inference Service 或 vLLM 部署。
正文节选
--- pipeline_tag: feature-extraction tags: - embedding - qwen3 - jina-embeddings-v5 - feature-extraction - sentence-transformers - multimodal - vision - audio - vllm - video - image-feature-extraction - audio-feature-extraction - video-feature-extraction - sentence-similarity language: - multilingual inference: false license: cc-by-nc-4.0 library_name: transformers --- <br><br> <p align="center"> <img src="https://huggingface.co/datasets/jinaai/documentation-images/resolve/main/logo.webp" alt="