Voyage AI Embeddings

Commercial Embedding Model API V AI Processing & RAG

Basic Information

  • Company/Brand: Voyage AI (acquired by MongoDB)
  • Country/Region: USA
  • Official Website: https://www.voyageai.com
  • Documentation: https://docs.voyageai.com
  • Type: Commercial Embedding Model API
  • Acquisition: Acquired by MongoDB in 2024
  • Latest Version: Voyage 4 Series (released in January 2026)

Product Description

Voyage AI is an AI company focused on embedding models. After being acquired by MongoDB in 2024, it became a core component of MongoDB's AI search ecosystem. The Voyage 4 series is its latest product, introducing the MoE (Mixture of Experts) architecture for the first time in production-level embedding models. Its innovative "shared embedding space" design allows models of different sizes to generate compatible embeddings, enabling switching between accuracy and speed without re-indexing.

Core Features/Characteristics

  • MoE Architecture: The first to adopt MoE architecture in production-level embedding models
  • Shared Embedding Space: Embeddings generated by all Voyage 4 models are compatible within the same vector space
  • Use voyage-4-large for document embedding to achieve the highest accuracy
  • Use voyage-4-lite for queries to achieve the fastest speed
  • No need for re-indexing
  • Domain-Specific Models: Offers domain-specific models for code (voyage-code-3), finance (voyage-finance-2), law (voyage-law-2), etc.
  • Multimodal Support: voyage-multimodal-3.5 supports multimodal embeddings for text and images
  • Input Type Optimization: Supports the input_type parameter to distinguish between queries and documents
  • Custom Dimensions: Supports configurable output dimensions

Model Matrix

ModelCharacteristicsPrice(/MTok)
voyage-4-largeHighest accuracy-
voyage-4Balanced version-
voyage-4-liteFastest speed-
voyage-3.5General-purpose$0.06
voyage-3.5-liteLightweight-
voyage-code-3Code-specific-
voyage-finance-2Finance-specific-
voyage-law-2Law-specific-
voyage-multimodal-3.5Multimodal-

Business Model

  • Free Quota:
  • voyage-3.5/3-large etc.: 200M tokens free
  • voyage-multilingual-2/finance-2/law-2 etc.: 50M tokens free
  • voyage-multimodal series: 200M text tokens + 150B pixels free
  • Pay-as-you-go: voyage-3.5 $0.06/MTok
  • Bulk Discount: Batch API offers a 33% discount
  • MongoDB Integration: Seamless integration with MongoDB Atlas

Target Users

  • MongoDB users and developers
  • Code search system developers
  • AI application teams in finance and legal industries
  • Developers requiring multimodal retrieval
  • Enterprise search and RAG system developers

Competitive Advantages

  • Unique and practical shared embedding space design
  • Domain-specific models (code, finance, law) excel in vertical fields
  • Highly competitive pricing ($0.06/MTok)
  • Generous free quota (200M tokens)
  • Deep integration with MongoDB ecosystem
  • MoE architecture provides efficient inference
  • Batch API offers a 33% discount

Limitations

  • Independence affected after acquisition by MongoDB
  • Closed-source commercial model, cannot be deployed locally
  • Increasingly tied to the MongoDB ecosystem
  • Fewer third-party integrations and community support compared to OpenAI

Relationship with the OpenClaw Ecosystem

Voyage AI provides a cost-effective embedding solution for OpenClaw, especially its $0.06/MTok pricing and 200M token free quota, which are very user-friendly for individual users. Domain-specific models (code, finance, law) can help OpenClaw provide more precise retrieval in specific fields. The shared embedding space design also allows OpenClaw to flexibly switch between accuracy and speed without reprocessing documents.

External References

Learn more from these authoritative sources: