MongoDB announced industry-first AI enhancements at MongoDB.local San Francisco. This integrates the core database with Voyage AI’s world-class embedding and re-ranking models to provide a unified data intelligence layer for production AI. By integrating these models directly into MongoDB’s platform infrastructure, developers can build and operate advanced applications at scale without moving or duplicating data, reducing the risk of illusions.
To support developers moving AI applications into production, MongoDB has introduced a set of new AI capabilities designed to simplify the way you build and operate intelligent applications. The company announced five embedding models for Voyage AI, a suite of embedding and retrieving models for MongoDB, automatic embedding for MongoDB Community Vector Search, embedding and reranking AI model APIs for Atlas, and an AI-powered data manipulation assistant for MongoDB Compass and Atlas Data Explorer. These capabilities strengthen MongoDB’s position as the leading AI-enabled data platform trusted by more than 60,000 customers running mission-critical workloads. Voyage AI models are available through MongoDB Atlas via API, integrated with the MongoDB community through managed automated embedding, and fully available as a standalone platform independent of MongoDB.
“The biggest challenge our customers face with AI is not experimentation, but ensuring that AI works at scale,” said Fred Roma, senior vice president of products and engineering at MongoDB. “Developers want fewer moving parts and a clearer path from prototype to production. With today’s release, MongoDB raises the bar and helps teams reduce complexity and focus on building AI applications that work in real-world, mission-critical environments.”
Transform data into AI intelligence
As projects move into production, many organizations are realizing that their existing data stacks were not designed to support context-aware, search-intensive workloads at scale. Developers must manage a fragmented mix of operational databases, vector stores, and model APIs, which introduces complexity, delay, and operational risk at the moments when speed and reliability are most important. This fragmentation is a major barrier to AI innovation and leads to real-world customer-facing impacts.
MongoDB addresses this problem by consolidating the core functionality needed to build and run AI applications in production into a single data platform. Instead of stitching together production databases, vector stores, and multiple pipelines, teams can keep production data and retrieval capabilities together, reducing latency and synchronization overhead. The result is AI applications with simpler architectures, faster iterations, and built to run reliably in production as well as demos. New features include:
- Cutting-edge accuracy with models from Voyage AI: The general availability of the new Voyage 4 series provides developers with high-performance embedded models that outperform Gemini and Cohere on public RTEB leaderboards, enabling more accurate searches at a lower cost. The Voyage 4 series includes the generic voyage-4.
Built-in models that balance acquisition accuracy, cost, and latency, the flagship voyage-4-large model for highest acquisition accuracy, voyage-4-lite for optimized latency and cost, and the openweight voyage-4-nano for local development and testing or on-device applications. - Facilitating context extraction from videos, images, and text: new voyage-multimodal-3.5 generally available
This model extends support for interleaved text and images to include video. Voyage AI’s voyage-multimodal-3 was the first production-grade embedded model to handle interleaved text and images, voyage-multimodal-3.5. advances this integrated processing approach to more effectively vectorize multimodal data to capture the most important semantic meaning from tables, graphics, figures, slides, PDFs, and more. This helps developers reduce the extensive effort required for complex document parsing, which can reduce search accuracy and reduce application reliability. - Automatic embedding for MongoDB Vector Search: Use Voyage AI to automatically generate and store high-fidelity embeddings every time data is inserted, updated, or queried. MongoDB eliminates the need for a separate embedding pipeline or external model service by handling embedding generation natively within the database. Embedding stays up-to-date as data changes, helping to maintain search accuracy and AI applications to maintain reliable context. The result is a simpler architecture with fewer moving parts, making it easier for teams to build and run AI-enabled applications in production. Automatic embedding is available in public preview with support in drivers (JavaScript, Python, Java, etc.) and AI frameworks such as LangChain and LangGraph (Python). Available today on MongoDB Community and coming soon to MongoDB Atlas.
“We were looking for a highly accurate embedded model, and Voyage AI provided accuracy at scale,” said Sudheesh Nair, Co-Founder and CEO of TinyFish. “The Python API that comes with Voyage out of the box is also very lightweight and very fast.”
“Today, companies need to move very fast, and very lean startups need to focus on what they are building,” said Rotem Weiss, CEO of Tavily. “MongoDB allows us to focus on what matters most: our customers and our business.”
For the first time, developers can build and run AI applications with operational data, semantic understanding, and retrieval on one system. MongoDB’s Atlas Embedding and Reranking API exposes Voyage AI models natively within Atlas, allowing teams to ship AI capabilities with enterprise-grade security, performance, and reliability infrastructure. MongoDB Compass and Atlas Data Explorer intelligent assistants are now generally available, providing natural language, AI-powered assistance with everyday data operations such as query optimization. MongoDB also introduced new AI skills certifications to help teams scale their data strategies, accelerate time to market, and reduce costs. This is the first of a broader set of AI skills planned for this year.
