$130M Series A to Build the Open Superintelligence Stack

Today, we're announcing that we've raised $130M, led by Radical Ventures, with participation from NVIDIA Ventures, Intel Capital, Dell Technologies Capital, and our existing investors. Bringing our total funding to over $150M to build the open superintelligence stack.

We're also joined by angels who are building the frontier themselves: John Schulman (Thinking Machines), Karim Atiyeh (Ramp), Aaron Levie (Box), Dwarkesh Patel, Milan Kovac (Tesla), Winston Weinberg (Harvey), Mike Knoop (Zapier, Ndea), Asher Spector (Flapping Airplanes), Jeff Wang (Cognition), Rohan Anil (Core Automation), Matthew Prince (Cloudflare), Brendan Foody (Mercor), Devansh Pandey (Standard Intelligence), Harrison Chase (Langchain), Nic Ouporov (Fleet) and many more.

Why now: RL changes who can build frontier AI

Pre-training concentrated frontier AI in a handful of labs. RL breaks that open: companies can now own their model optimization loop — train directly on their own product, optimize for their specific workflows, and build agents that improve continuously in production.

Owning this loop is how you build a compounding moat in the agentic era. The only missing piece has been the infrastructure — until now, it lived exclusively inside the labs.

The Open Superintelligence Stack

We train open frontier models and ship the same stack to our customers. It spans the full stack of training, deploying and continuously improving models — compute, large-scale RL, environments, sandboxes, evals, and deployment.

Who's building on Prime Intellect

We are grateful to over 6k customers working with us including many of the leading AI startups, neolabs and enterprises, who use our stack — across compute, RL and post-training, sandboxes, inference, environments, and evaluations.

In under a year, that demand has scaled to over $100m in annualized revenue.

Companies like Ramp beat closed frontier models using our post-training stack. Ramp trained a 35B model on Lab that beat Opus at spreadsheet search, running 27% faster and far cheaper than Haiku. Read the case study →

“We worked with Prime Intellect to train Fast Ask on Lab — a small RL-trained subagent that helps the Ramp Sheets agent find answers inside spreadsheets. The result beat the frontier models on accuracy while running at faster speeds and a fraction of the cost. Rather than wait on a better frontier model, we trained our own for the workflow that mattered to us”

Karim Atiyeh

Ramp Co-CEO

What's next

We're scaling every layer of the stack — ever-larger compute clusters, larger RL runs, and the stack for agentic training, inference and continual learning.

Beyond that, we are placing ambitious bets at the frontier of where the puck is going and build infrastructure for the problems we believe are most consequential, such as:

• Long-horizon agents and Recursive Language Models (RLMs). Today's models break down over long contexts; RLMs manage their own context and coordinate sub-agents. We've been scaling RLM training over the last months — we believe it will be the scaling paradigm for agents that work for days.
• Automate AI research and science, across all aspects from pre-training (autonomous nanogpt), to RL (General Agent) and beyond.
• Continual learning. The future is models that learn in production, where training and inference collapse into a single continuous loop. Our stack was built for this world — a tight integration between RL rollouts, training, and serving.

Join us

We're a small team racing the most well-funded closed labs in the world to build superintelligence in the open. The same stack that trains our frontier models is now in the hands of thousands of teams.

If you want to train frontier models you own — on your data, your workflows, your product — we'd love to build with you. Start training →

If you want to build the infrastructure of open superintelligence, we're hiring across RL, inference, distributed systems, and compute. See open roles →