Introducing the Embedding and Reranking API on MongoDB Atlas

The next frontier for AI isn’t simply more capable models. It’s better context. As LLMs become embedded in every process and product, their accuracy and trust depend on grounding generation in the right data. Search and retrieval are foundational to this shift, powering everything from AI chatbots and assistants to fully autonomous agents. But building AI retrieval today means stitching together databases, vector search, and retrieval model providers—each introducing operational complexity.

To address this, we’re introducing the Embedding and Reranking API on MongoDB Atlas, now in public preview. This service gives developers direct access to Voyage AI’s state-of-the-art retrieval models within the Atlas ecosystem. Use it as a standalone API with any stack or combine it with MongoDB Atlas features like Vector Search to build complete retrieval pipelines on a single, enterprise-grade platform.

Bringing frontier AI retrieval models to MongoDB Atlas

Embedding and reranking models are core components of modern retrieval systems, driving the accuracy of recommendation, search, and RAG applications.

Yet selecting the right model is often overlooked. Results from the Retrieval Embedding Benchmark (RTEB) show that retrieval accuracy varies significantly across tasks, with performance gaps of up to nearly 50% between the strongest and weakest models. 

Choosing the right retrieval models is critical for:

• Improving retrieval accuracy

• Minimizing storage costs, latency, and context bloat

• Addressing the needs of specific industries, such as law and finance.

Voyage AI offers a comprehensive suite of state-of-the-art embedding and reranking models that consistently deliver the highest accuracy across categories on RTEB. The text embedding models support flexible dimensionality, ranging from 256 to 2048, as well as quantization, giving developers control over the quality-cost-latency tradeoff. Beyond general-purpose embeddings, Voyage offers models tailored for domain-specific applications (such as law, finance, and code), full-document understanding, multimodal use cases, and high-precision reranking for two-stage retrieval pipelines.

Voyage AI’s models are now natively available on MongoDB Atlas via the new Embedding and Reranking API, adding to the AI retrieval capabilities available through our platform. This serverless API offers simple token-based pricing and is open for use with any stack, not just MongoDB.

Today, we are also launching the Voyage 4 model series, our most advanced embedding models yet. This new generation of models features an industry-first shared embedding space, making the models compatible with one another and giving developers more flexibility to optimize for accuracy, speed, and cost. You can now start building with the Voyage 4 series using the Embedding and Reranking API on MongoDB Atlas.

MongoDB Atlas: A unified platform for AI retrieval

MongoDB Atlas now offers the foundational operational database, advanced vector search capabilities, and state-of-the-art retrieval models needed to build AI applications on a single platform. Teams get consolidated and unified user management, monitoring, and billing, backed by enterprise-grade security and controls already trusted by tens of thousands of customers, including over 75% of Fortune 100 companies.

This matters for building production AI systems. Scaling LLM applications requires delivering the right context at the right time, which means tightly integrating operational data with high-performance search. MongoDB Atlas's industry-leading document database and built-in vector search capabilities already power retrieval systems across AI startups and large enterprises. The Embedding and Reranking API completes these capabilities, enabling teams to build end-to-end AI retrieval from data storage to vector search to embeddings and reranking entirely within MongoDB Atlas.

Better AI starts with better context. With Voyage AI's state-of-the-art models now accessible via the Embedding and Reranking API on MongoDB Atlas, teams can build complete retrieval pipelines on a proven platform—from data storage and vector search to embeddings and reranking. The API is available now in public preview.