MongoDB announces new product innovations in AI stack

MongoDB has announced a range of product innovations and AI partner ecosystem expansions that make it faster and easier for customers to build accurate, trustworthy, and reliable AI applications at scale. By providing industry-leading embedding models and a fully integrated, AI-ready data platform—and by assembling a world-class ecosystem of AI partners—MongoDB is giving organizations everywhere the tools to deliver reliable, performant, cost-effective AI.

MongoDB continues to invest in streamlining the AI stack and introducing more performant, more cost-effective models. Customers can integrate Voyage AI’s latest embedding and reranking models with their MongoDB database infrastructure. MongoDB has also increased its interoperability with industry-leading AI frameworks—by launching the MongoDB MCP Server to give agents access to tools and data, and by expanding its comprehensive AI partner ecosystem to give developers more choice.

These capabilities are fueling substantial momentum among developers building next-generation AI applications. Enterprise AI adopters like Vonage, LGU+, and The Financial Times—plus approximately 8,000 startups, including the timekeeping startup Laurel, and Mercor, which uses AI to match talent with opportunities—have chosen MongoDB to help build their AI projects in just the past 18 months. Meanwhile, more than 200,000 new developers register for MongoDB Atlas every month.

“Databases are more central than ever to the technology stack in the age of AI. Modern AI applications require a database that combines advanced capabilities—like integrated vector search and best-in-class AI models—to unlock meaningful insights from all forms of data (structure, unstructured), all while streamlining the stack,” said Andrew Davidson, SVP of Products at MongoDB. “These systems also demand scalability, security, and flexibility to support production applications as they evolve and as usage grows. By consolidating the AI data stack and by building a cutting-edge AI ecosystem, we're giving developers the tools they need to build and deploy trustworthy, innovative AI solutions faster than ever before.”

Voyage AI by MongoDB recently introduced industry-leading embedding models designed to unleash new levels of AI accuracy at a lower cost –

●        Context-aware embeddings for better retrieval: The new voyage-context-3 model brings a breakthrough in AI accuracy and efficiency. It captures the full document context—no metadata hacks, LLM summaries, or pipeline gymnastics needed—delivering more relevant results and reducing sensitivity to chunk size. It works as a drop-in replacement for standard embeddings in RAG applications.

●        New highs in model performance: The latest general-purpose models, voyage-3.5 and voyage-3.5-lite, raise the bar on retrieval quality, delivering industry-topping accuracy and price-performance.

●        Instruction-following reranking for improved accuracy: With rerank-2.5 and rerank-2.5-lite, developers can now guide the reranking process using instructions, unlocking greater retrieval accuracy. These models outperform competitors across a comprehensive set of benchmarks.

MongoDB also recently introduced the MongoDB Model Context Protocol (MCP) Server in public preview. This server standardizes connecting MongoDB deployments directly to popular tools like GitHub CoPilot in Visual Studio Code, Anthropic's Claude, Cursor, and Windsurf—allowing developers to use natural language to interact with data and manage database operations—and streamlines AI-powered application development on MongoDB, accelerating workflows, boosting productivity, and reducing time to market.

Since launching in public preview, the MongoDB MCP Server has rapidly grown in popularity, with thousands of users building on MongoDB every week. MongoDB has also seen significant interest from large enterprise customers looking to incorporate MCP as part of their agentic application stack.

“Many organizations struggle to scale AI because the models themselves aren’t up to the task. They lack the accuracy needed to delight customers, are often complex to fine-tune and integrate, and become too expensive at scale,” said Fred Roma, SVP of Engineering at MongoDB. “The quality of your embedding and reranking models is often the difference between a promising prototype and an AI application that delivers meaningful results in production. That’s why we’ve focused on building models that perform better, cost less, and are easier to use—so developers can bring their AI applications into the real world and scale adoption.”