MongoDB acquires Voyage AI to reduce hallucinations in AI applications

Database engine developer  MongoDB has acquired Voyage AI in order to help enterprises reduce hallucinations in AI-powered applications.

“With this acquisition, MongoDB will help deliver AI-powered, never-before-seen high quality outputs (via high quality data retrieval) as a fully formed database solution — not with tacked on additions,” the database developer firm said in a statement.

Voyage AI develops embedding and reranking models used in retrieval-augmented generation (RAG). These models will help reduce hallucinations by enhancing vector search with new embeddings that better capture meaning and improving retrieval accuracy by refining search results to a particular query, MongoDB explained.

Voyage AI is already used by Anthropic, LangChain, Harvey, and Replit.

Additionally, integrating Voyage AI’s models will allow MongoDB enterprise customers to create domain-specific AI with fine-tuned models optimized for different industries such as financial services, healthcare, and law, and use cases such as code generation, the company said.

How enterprise developers benefit

Voyage AI’s integration with MongoDB will ensure that developers no longer need to manage external embedding APIs, standalone vector stores, or complex search pipelines, Dev Ittycheria, CEO of MongoDB, wrote in a blog post.

“AI retrieval will be built into the database itself, making semantic search, vector retrieval, and ranking as seamless as traditional queries,” Ittycheria explained.

For enterprises, the acquisition translates to faster time-to-value and greater confidence in scaling AI applications, he said.

The integration itself will happen in three phases with Voyage AI’s text embedding, multi-modal embedding, and reranking models remaining widely available through Voyage AI’s current APIs and via the AWS and Azure Marketplaces in the first phase.

In the second phase, MongoDB will embed Voyage AI’s capabilities into MongoDB Atlas, starting with an auto-embedding service for Vector Search, which will handle embedding generation automatically.

“Native reranking will follow, allowing developers to boost retrieval accuracy instantly,” Ittycheria wrote.

In the third phase, MongoDB will add advanced AI-powered retrieval with enhanced multi-modal capabilities.

“We also plan to introduce instruction-tuned models, allowing developers to refine search behavior using simple prompts instead of complex fine-tuning,” Ittycheria explained, adding that it will be complemented by embedding lifecycle management in MongoDB Atlas optimization of AI applications.

The value of the transaction, according to a Bloomberg report, was $220 million.