OpenAI entered into a definitive agreement to acquire Polish neptune.ai, a metrics dashboard platform founded in 2017 that tracks experiments and monitors training for frontier AI models.
Neptune will integrate its experiment-tracking system deep into OpenAI’s training stack, enabling researchers to compare thousands of runs and surface issues in real time. The acquisition brings Neptune’s eight-year focus on model development workflows directly into one of the world’s leading AI research organizations.
Neptune’s killer metrics on how models learn
Neptune launched in 2017, the year the transformer architecture paper emerged. Since then, the company concentrated on supporting researchers during what founder and CEO Piotr Niedźwiedź describes as “the iterative, messy, and unpredictable phase of model training.”
The platform gives teams visibility into complex training workflows, allowing them to analyze metrics across layers and monitor how models evolve. Or, as CEO himself wrote on c0mpany’s blog, the best description of Neptune’s core function comes from Szymon Sidor, who spent nearly a decade at OpenAI:
“OpenAI research converts compute into understanding. At the interface of compute and understanding are metrics. Neptune is a metrics dashboard company.”
OpenAI and Neptune worked closely before the acquisition to develop tools tailored for foundation model development. That collaboration will now deepen as Neptune’s team joins OpenAI to build training infrastructure at scale.
Integration plans and service wind-down
Jakub Pachocki, OpenAI’s Chief Scientist, explained the technical rationale: “Neptune has built a fast, precise system that allows researchers to analyze complex training workflows. We plan to iterate with them to integrate their tools deep into our training stack to expand our visibility into how models learn.”
Neptune will wind down its external services over the next few months. The company committed to working with existing customers and users to smooth the transition. Neptune’s platform currently enables researchers to track experiments, compare thousands of runs, and debug model behavior as training unfolds.
Niedźwiedź framed the move as an opportunity to operate at a different scale: “This is an exciting step for us. We’ve always believed that good tools help researchers do their best work. Joining OpenAI gives us the chance to bring that belief to a new scale.”
Neptune’s team will continue building tools to monitor, debug, and evaluate frontier models within OpenAI, supporting the organization’s research as it works toward artificial general intelligence. The acquisition remains subject to standard closing conditions.
