Prime Intellect is Building the Open Superintelligence Stack

Prime Intellect, making frontier AI training accessible to every company, led by Vincent Weisser and Johannes Hagemann, has raised $130M, led by Radical Ventures, with participation from NVIDIA Ventures, Intel Capital, Dell Technologies Capital, and existing investors, bringing the total funding to over $150M to build the open superintelligence stack.

Reinforcement learning (RL) is reshaping how frontier AI systems are built. While pre-training concentrated advanced AI development among a small number of labs, RL enables companies to optimize models using their own data, products, and workflows, creating agents that continuously improve in production.

Owning this optimization loop could become a key competitive advantage in the agentic AI era. Until now, the infrastructure required to build and run these systems has largely remained within frontier AI labs.

The Open Superintelligence Stack enables the training, deployment, and continuous improvement of open frontier models. The platform provides a full-stack infrastructure layer covering compute, large-scale reinforcement learning, environments, sandboxes, evaluations, and deployment.

Prime Intellect is used by more than 6,000 customers, including leading AI startups, research labs, and enterprises, leveraging its stack across compute, reinforcement learning and post-training, sandboxes, inference, environments, and evaluations.

In less than a year, demand for the platform has grown to more than $100 million in annualized revenue.

Karim Atiyeh said Prime Intellect enabled Ramp to train a specialized RL-powered subagent for spreadsheet workflows, achieving higher accuracy than frontier models while operating faster and at lower cost. The experience demonstrated the ability for companies to build and optimize their own models for specific use cases rather than relying solely on general-purpose frontier models.

Prime Intellect is scaling every layer of its AI infrastructure stack, including larger compute clusters, advanced reinforcement learning runs, and tools for agentic training, inference, and continual learning.

The company is also investing in emerging AI paradigms, including long-horizon agents and Recursive Language Models (RLMs) designed to handle extended tasks, automated AI research and science, and continual learning systems where models improve continuously through integrated training and deployment loops.