Voyage AI

What it is for

Voyage AI is an embedding-and-rerank API: turn text (or images/video) into vectors, then optionally rerank the hits before they reach a generator. Use it when the job is retrieval quality you can meter in tokens, not when you need a vector database or a chat seat.

What makes it unique: the cover is the official voyage-ai/voyageai-python GitHub file list — voyageai/, MIT in the About pane, README “embedding and rerankers” — not a pricing card.

Key features

• Text embeddings — vo.embed(...) on the voyage-4 family (voyage-4-lite / voyage-4 / voyage-4-large, plus code/finance/law SKUs) with input_type query or document (Quickstart; Pricing). One embed call; those names are SKU, not five products.
• Multimodal embeddings — the multimodal endpoint mixes text tokens with image/video pixels (voyage-multimodal-3.5 / 3) so a photo and a caption share one vector space (pricing Multimodal). A different endpoint, not a text-only flag.
• Rerankers — vo.rerank(query, documents, model="rerank-3") scores an already-retrieved pile (Quickstart “Refinement with rerankers”). A second pass, not a first embed.
• Batch + Files API — submit many embed/rerank jobs with a 12-hour window at a 33% discount; Files stores the upload for 30 days at $0.05/GB/mo (pricing Batch and Files). A queue, not a live call.
• Fine-tuned dedicated models — custom embeds on reserved instances, quoted by sales (contact@voyageai.com). A contract box, not the public token table.

The Python client and VOYAGE_API_KEY (now also minted in MongoDB Atlas) are install / SDK. Tokens (and pixels on multimodal) are the meter. Model names and Matryoshka dimensions are SKU.

Who it's for

Best fit: RAG teams who will burn the 200 million free tokens on voyage-4 / voyage-4-lite and then pay $0.02–$0.12 per million, not a team looking for a $20 chat seat.

Pricing

Verified 2026-10-22 (Asia/Shanghai) from docs.voyageai.com/docs/pricing. Usage-based; billed to the account on the API key. There is no seat card.

• Free tokens — first 200 million tokens on voyage-4-large, voyage-4, voyage-4-lite, voyage-context-4, and voyage-code-4. First 50 million on voyage-finance-2 / voyage-law-2 / voyage-code-2. Multimodal: first 200M text tokens and 150B pixels.
• Text embed — voyage-4-lite $0.02 / 1M tokens; voyage-4 $0.06; voyage-4-large / voyage-context-4 / voyage-code-4 $0.12. Specialized finance/law/code-2: $0.12 after 50M free.
• Multimodal — $0.12 / 1M text tokens + $0.60 / 1B pixels. Per-image floor $0.00003, ceiling $0.0012.
• Rerank — rerank-3 $0.05 / 1M tokens; rerank-3-lite $0.02. First 200 million tokens free.
• Batch — 33% off the live rates. Free-token credits do not apply to Batch.
• Files storage — $0.05 / GB / month; 30-day retention.
• Fine-tune / dedicated — custom (contact@voyageai.com).

Older voyage-3.x rows have no free tokens. Confirm checkout if MongoDB Atlas bills the same key.

Limitations / costs

• No public seat. Dollars are per million tokens (and pixels).
• Batch does not spend the free-token pot. Live calls do.
• Embed API, not a vector store. You still need Qdrant, Chroma, or Atlas to keep the vectors.

Alternatives

• Cohere — another hosted embed/rerank API when Voyage’s voyage-4 table is not the job.
• Qdrant — the store + optional Cloud Inference when you want vectors to live next to search.
• OpenRouter — a model gateway when you need many providers, not a specialist embedder.

Official tutorial

Official MongoDB for Developers channel video: Which Embedding Model should You Use for AI Applications? | MongoDB Voyage AI Quickstart (oEmbed author: MongoDB for Developers). Voyage is MongoDB’s embedding line; this is the first-party getting-started.

Written path: Quickstart Tutorial plus Pricing and the Atlas Voyage Quick Start.

Website

https://www.voyageai.com

Repository

https://github.com/voyage-ai/voyageai-python

Source

Catalog row curated for AI Tools Nav. Cover is the official voyage-ai/voyageai-python GitHub file list (MIT + voyageai/). Dollars checked on docs.voyageai.com/docs/pricing (voyage-4-lite $0.02; voyage-4 $0.06; voyage-4-large $0.12 / 1M; 200M free tokens). Tutorial oEmbed author: MongoDB for Developers.

Last verified

2026-10-22 (Asia/Shanghai) editorial check.