MongoDB unveils enterprise-focused AI models
The vendor's new models enhance document context capture and enable developers to guide the reranking process with instructions, while improving accuracy.
Unified database platform provider MongoDB unveiled new AI models this week, seeking to address some of the challenges enterprises face when adopting AI technologies.
On Monday, the vendor introduced voyage-3.5 and voyage-3.5-lite, two general purpose models, rerank-2.5 and rerank-2.5-lite, which enables developers to guide the reranking process with instructions and better retrieval accuracy. The vendor also introduced voyage-3-context, a model that can capture full document context and acts as a drop-in replacement for embeddings or the act of putting one system into another to create a seamless user application.
The introduction of the voyage- and rerank models reflects a new trend within the AI market. Instead of focusing on general purpose models and consumer products like ChatGPT, there is a shift to ensure that the models are customized for the enterprise.
"Enterprises need more than consumers require," said Jason Andersen, an analyst at Moor Insights & Strategy. "They need their AI to be more accurate. They need it to be customized to meet the needs of their branding or their guidelines."
Customized models
Because of this need, data is becoming increasingly prominent in the enterprise AI conversation. MongoDB is trying to introduce AI capabilities such as embedding vectors into conversations about data and AI, Andersen said.
"It's bringing the two worlds together," he said, adding that it's a good combination for enterprises that want more out of AI technology than just models.
MongoDB's experience in data management and its history with open source give it an inherent advantage in supporting developers work with all kinds of data. Furthermore, compared with its database competitors, MongoDB also has fewer applications within its product suite to support, Andersen said.
Moreover, given the vendor's credibility in the NoSQL space and the new models, "they've integrated the model and the embedding tools into the database infrastructure itself instead of as a separate thing," he added.
MongoDB shows the simplicity of its AI stack through the embedding tools it introduced, said Stephen Catanzano, an analyst at Omdia, a division of Informa TechTarget.
"By integrating the best AI models and vectors directly with the best database for search at its core, you have a best-of-breed solution with zero metadata hacks, LLM summaries or bespoke pipeline logic," Catanzano said. "I like the combination of MongoDB with the integrations. It's what the market is looking for."
Some Challenges
However, Andersen said MongoDB faces a crowded market, which will be a challenge. Vendors including Oracle, IBM and AWS also have robust data offerings.
Moreover, enterprises are hitting a plateau after being flood by AI technology possibilities that might not be delivering enough of an ROI.
"Maybe they've done some things with AI, and it worked pretty well; or they've done it, and [now they're] trying to think about what the next big business challenge I can solve with AI is," Andersen said. "But maybe it's not quite accurate enough, the performance isn't as good or the cost might be too high."
He said that AI in the enterprise is receiving a high degree of scrutiny, so vendors need to educate enterprises on the next steps in AI technology.
"A company like MongoDB can help with that because they have a lot of experience in data, and again, data is the foundation of great AI," Andersen said.
MongoDB added new partners to its ecosystem, including Galileo and Temporal, along with the latest models. Galileo is a AI reliability and observability platform. With its availability on MongoDB’s platform, it allows AI applications and agents built on MongoDB to be deployed with continuous evaluations and monitoring. Temporal is an open source platform that developers can use to orchestrate AI use cases built on MongoDB.
A new MongoDB Model Context Protocol Server connects MongoDB directly to AI tools including GitHub, Copilot, Claude and Windsurf. The capability is now in preview.
Esther Shittu is an Informa TechTarget news writer and podcast host covering artificial intelligence software and systems.
