Model list-OpenSearch(Open Search)-阿里云帮助中心

Vector Search Edition supports built-in models, AI Search Open Platform models, and user-defined models. Use the Model list page to view and manage them.

AI Search Open Platform models

You can call AI Search Open Platform models to pre-process raw images or text for text-to-image search, image-to-image search, and semantic search. For more information, see Call AI Search Open Platform model services.

Model ID	Model name	Model class	Model description
ops-text-embedding-001	OpenSearch universal text embedding service-001	Text embedding	Provides a multilingual (40+) text embedding service. The maximum input text length is 300. The output has 1536 vector dimensions.
ops-text-embedding-zh-001	OpenSearch text embedding service-Chinese-001	Text embedding	Provides a Chinese text embedding service. The maximum input text length is 1024. The output has 768 vector dimensions.
ops-text-embedding-en-001	OpenSearch text embedding service-English-001	Text embedding	Provides an English text embedding service. The maximum input text length is 512. The output has 768 vector dimensions.
ops-text-embedding-002	OpenSearch universal text embedding service-002	Text embedding	Provides a multilingual (100+) text embedding service. The maximum input text length is 8192. The output has 1024 vector dimensions.
ops-text-sparse-embedding-001	OpenSearch text sparse vector service-001	Text sparse vectorization	Provides a multilingual (100+) text embedding service. The maximum input text length is 8192.
ops-image-analyze-ocr-001	Image Text Recognition Service	Image content parsing	Detects and extracts text from images by using OCR. Applicable to scenarios such as image retrieval and Q&A pairs.
ops-image-analyze-vlm-001	Image Content Recognition Service	Image content parsing	Parses and recognizes text in images by using a multimodal Large Language Model (LLM). Applicable to scenarios such as image retrieval and Q&A pairs.

User-defined models

To use your own models, add them in the console and then add them to the configuration table. For more information, see User-defined models.

Built-in models

Model ID	Model name	Model class	Model description
clip	Universal image-to-vector model-512 dimensions	Image vectorization	A universal image-to-vector model that supports image-to-image search and text-to-image search. The output has 512 vector dimensions.
clip_ecom	E-commerce enhanced image-to-vector model-512 dimensions	Image vectorization	An image-to-vector model enhanced for E-commerce scenarios that supports image-to-image search and text-to-image search. The output has 512 vector dimensions.
ops-text-embedding-1024-000-20231001	Enhanced Chinese text-to-vector-768 dimensions	Text embedding	Supports Chinese text-to-vector conversion. The Chinese text length cannot exceed 1024 tokens. The output has 768 vector dimensions.
ops-text-embedding-512-000-20231001	Enhanced Chinese text-to-vector-768 dimensions	Text embedding	Supports Chinese text-to-vector conversion. The Chinese text length cannot exceed 512 tokens. The output has 768 vector dimensions.
ops-text-embedding-128-000-20231001	Enhanced Chinese text-to-vector-768 dimensions	Text embedding	Supports Chinese text-to-vector conversion. The Chinese text length cannot exceed 128 tokens. The output has 768 vector dimensions.
ops-text-embedding-512-en-000-20231001	Enhanced English text-to-vector-768 dimensions	Text embedding	Supports English text-to-vector conversion. The English text length cannot exceed 512 tokens. The output has 768 vector dimensions.
ops-text-embedding-128-en-000-20231001	Enhanced English text-to-vector-768 dimensions	Text embedding	Supports English text-to-vector conversion. The English text length cannot exceed 128 tokens. The output has 768 vector dimensions.