MongoDB believes that improved data retrieval, rather than simply scaling up AI models, is crucial for building trustworthy enterprise AI systems. As agentic systems and Retrieval-Augmented Generation (RAG) gain traction in production environments, the database provider has identified retrieval quality as a significant bottleneck that can negatively impact accuracy, cost-efficiency, and user confidence, even when the underlying AI models are robust.

To address this challenge, MongoDB recently launched four new versions of its embeddings and reranking models, collectively known as Voyage 4. These models are designed to enhance the precision and effectiveness of data retrieval processes within AI applications. The Voyage 4 family includes voyage-4 embedding, a general-purpose model; voyage-4-large, considered MongoDB's flagship model; voyage-4-lite, optimized for low-latency and cost-sensitive applications; and voyage-4-nano, intended for local development, testing, and on-device data retrieval. Voyage-4-nano marks MongoDB's first foray into open-weight models.

All Voyage 4 models are accessible through an API and on MongoDB's Atlas platform. According to the company, these models outperform comparable offerings in the market.

The emphasis on retrieval quality highlights a growing awareness within the AI community that the performance of agentic and RAG systems is heavily dependent on the ability to efficiently and accurately access relevant information. RAG systems, in particular, rely on retrieving information from a knowledge base to augment the prompts given to large language models (LLMs), thereby improving the accuracy and reliability of the generated responses. Poor retrieval can lead to inaccurate or incomplete information being fed to the LLM, resulting in flawed outputs and diminished user trust.

By focusing on optimizing embeddings and reranking models, MongoDB aims to improve the overall performance and trustworthiness of AI applications that rely on efficient data retrieval. The availability of different Voyage 4 models caters to a range of use cases, from general-purpose applications to resource-constrained environments. The open-weight nature of Voyage-4-nano also encourages community collaboration and innovation in the field of data retrieval for AI.

The development signifies a shift in focus within the AI landscape, acknowledging that advancements in model size alone are not sufficient to guarantee reliable and accurate AI systems. Instead, a holistic approach that considers the entire AI pipeline, including data retrieval, is essential for building trustworthy and effective enterprise AI solutions. The company said the models are available now.