Voyage AI
An AI company specializing in text embedding and reranking models. It provides high-quality vector representations for use cases including code retrieval, long-document understanding, and multilingual scenarios.
Core Facts
@ai-sdk/voyage is the official provider package within the Vercel AI SDK dedicated to Voyage AI embedding models
Voyage AI is known for its high-quality text embedding models, which are widely used in RAG, semantic search, and vectorization scenarios
Voyage AI is a company focused on text embedding and reranking models
Timeline (last 90 days)
嵌入模型是将离散输入映射为稠密向量的神经网络模型，代表有OpenAI的text-embedding-ada-002、Voyage AI的voyage系列，以及开源的BGE、E5、GTE等。
Voyage AI博客给出了完整write-up，代码已在GitHub开源（mongo-predict）。
Voyage AI工程师在Reddit发帖称，使用嵌入模型加sklearn分类器的方案比决策模型Jev错误率降低了37%。
What truly affects retrieval performance is the model version, the vector dimensions, and whether call parameters such as inputType (e.g. document/query) are used.
For projects not yet integrated, this update itself is not a reason to choose the library; what is more worth attention is the Voyage model's performance on embedding quality and cost.
Voyage AI's models such as voyage-3 belong to the embedding model category.
Dependency updates do not change the capabilities or billing of the Voyage AI model; what actually affects output quality is the model parameter specified in the request and the input content.
Voyage AI's own model service (such as the voyage-3 series) and the npm package @ai-sdk/voyage are two different things: the former is a remote API, while the latter is merely an SDK adapter layer for calling it.
In AI SDK, @ai-sdk/voyage wraps the Voyage AI service into an interface consistent with providers such as OpenAI and Cohere, enabling developers to call different vendors' vector models with a unified embed/embedMany
Voyage AI's models include the voyage-3, voyage-code-3 and other series, and are commonly used in scenarios such as semantic retrieval, RAG, clustering, and deduplication
11 more timeline events
All Facts (20)
@ai-sdk/voyage is the official provider package within the Vercel AI SDK dedicated to Voyage AI embedding models
90%VerifiedVoyage AI is known for its high-quality text embedding models, which are widely used in RAG, semantic search, and vectorization scenarios
75%VerifiedVoyage AI is a company focused on text embedding and reranking models
75%UnverifiedVoyage has embedding model versions specifically optimized for vertical domains such as code, legal, and medical.
70%UnverifiedVoyage's models can be called via its REST API, and @ai-sdk/voyage wraps that call into a provider conforming to the unified interface of the Vercel AI SDK.
70%Unverified嵌入模型是将离散输入映射为稠密向量的神经网络模型，代表有OpenAI的text-embedding-ada-002、Voyage AI的voyage系列，以及开源的BGE、E5、GTE等。
50%UnverifiedVoyage AI工程师在Reddit发帖称，使用嵌入模型加sklearn分类器的方案比决策模型Jev错误率降低了37%。
50%UnverifiedVoyage AI博客给出了完整write-up，代码已在GitHub开源（mongo-predict）。
50%UnverifiedWhat truly affects retrieval performance is the model version, the vector dimensions, and whether call parameters such as inputType (e.g. document/query) are used.
50%UnverifiedFor projects not yet integrated, this update itself is not a reason to choose the library; what is more worth attention is the Voyage model's performance on embedding quality and cost.
50%UnverifiedVoyage AI's models such as voyage-3 belong to the embedding model category.
50%UnverifiedVoyage AI's own model service (such as the voyage-3 series) and the npm package @ai-sdk/voyage are two different things: the former is a remote API, while the latter is merely an SDK adapter layer for calling it.
50%UnverifiedDependency updates do not change the capabilities or billing of the Voyage AI model; what actually affects output quality is the model parameter specified in the request and the input content.
50%UnverifiedVoyage AI's models include the voyage-3, voyage-code-3 and other series, and are commonly used in scenarios such as semantic retrieval, RAG, clustering, and deduplication
50%UnverifiedIn AI SDK, @ai-sdk/voyage wraps the Voyage AI service into an interface consistent with providers such as OpenAI and Cohere, enabling developers to call different vendors' vector models with a unified embed/embedMany
50%UnverifiedVoyage AI is a service provider focused on delivering high-quality embedding models, and its models are widely used in scenarios such as RAG and semantic search.
50%UnverifiedVoyage AI has trained dedicated embedding models for specific domains such as code, law, and finance, which often outperform general-purpose embedding models in retrieval accuracy for vertical scenarios.
50%UnverifiedVoyage AI is a company focused on text embedding models, with core products including the voyage-2 and voyage-large-2 series.
50%UnverifiedVoyage AI's embedding models stand out in standard benchmarks such as MTEB, and are particularly strong in code retrieval and long-document understanding scenarios.
50%UnverifiedVoyage AI's embedding models perform notably in vertical domains such as code retrieval, legal, and finance, and support a context window of up to 32K tokens.
50%