As recorded at collection: 200 million free tokens per flagship Voyage AI model
As recorded at collection: Voyage AI is an API service that turns the meaning of text into numbers for use in search. It offers new accounts 200 million free tokens per flagship model. You can sign up and start using it right away without registering a credit card, and there is no usage expiration date.
- What you get
- 200 million free tokens per flagship model · An additional 150 billion pixels for multimodal models
- Price
- Free · 200 million tokens per model are worth about $4 at OpenAI's text-embedding-3-small rates
- Valid until
- No time limit
- Requirements
- Free limits apply separately to each model on new accounts · Sign up with a MongoDB Atlas account · No credit card registration or separate database creation required · No expiration date
Details
Voyage AI is an API service that turns text or images into numbers to find semantically similar materials and reorder search results. It was acquired by MongoDB in early 2024 and is now called 'Voyage AI by MongoDB'. New accounts receive 200 million free tokens for each flagship model (voyage-4-large, voyage-4, voyage-4-lite, voyage-context-3, voyage-code-3, rerank-2.5, rerank-2.5-lite). Older specialized models offer 50 million tokens per model, while multimodal models that process text and images together provide a separate limit of 200 million text tokens and 150 billion pixels.
Sign up for Voyage AI with a free MongoDB Atlas account, then issue an API key (an authentication string required to use the service) to get started. There is no need to register a credit card or create a separate database. The signup process is rated 1 out of 5 for difficulty, meaning it is described as ready to use immediately.
The free limit applies separately to each model on every account, so you can try several models at once, and no separate expiration date has been announced. The introductory post describes this as one of the largest free limits currently available for searching materials, answering questions based on documents, or searching code, and says its performance is comparable to Cohere or OpenAI.
Sources and verification history
- Collection channel
- aicredits
- Collected at
- July 25, 2026, 16:43 KST
- First published
- Updated
- Updates
- July 26 23:52 · Reclassified due to a change in criteria