Voyage AI

Search-foundation APIs — Reader, embeddings, and reranker — for grounding LLMs.

A suite of search-foundation APIs for retrieval and RAG: a Reader that turns any URL or web search into LLM-ready markdown, multilingual multimodal embeddings, and a reranker. One key spans every service, the Reader is open source, and the embedding models are also released as open weights for self-hosting.

Managed multimodal search over your text, PDFs, images, and video.

A fully managed search engine that indexes text, PDFs, tables, images, and video across 100+ languages without hand-tuning embeddings or a multi-stage pipeline. The Berlin team is best known for its open-source mxbai embedding and reranking models, which the hosted product builds on. Access it via dashboard, Python/TypeScript SDKs, or MCP.

Enterprise-grade LLMs, embeddings, and retrieval built for private deployment.

Cohere builds large language models for the enterprise rather than the consumer. Its Command models cover agentic, multimodal, and multilingual generation; Embed and Rerank power high-quality search and retrieval; Aya is a multilingual research family spanning 70+ languages; and North is a workplace AI platform built on top. Cohere's emphasis is data control — models can run in a private VPC, on-premises, or via a managed Model Vault.