# Prime Intellect

> Prime Intellect builds infrastructure for training, evaluating and deploying AI agents. It began as a marketplace for distributed GPU compute and now offers Lab, open RL frameworks, shared environments, hosted evaluations, inference and sandboxes.

- **Founded:** 2024
- **Headquarters:** San Francisco, United States
- **Founders:** Vincent Weisser (Co-founder and CEO), Johannes Hagemann (Co-founder and CTO)
- **Products:** Prime Intellect Compute, PRIME and prime-rl, Environments Hub and verifiers, Prime Intellect Lab, Prime Sandboxes
- **Notable:** Trained INTELLECT-1, a 10B model, across five countries and three continents on up to 112 H100 GPUs., Released INTELLECT-2, a 32B model trained through globally distributed reinforcement learning., Reported more than 10,000 Lab training jobs during its 2026 beta.

## Products & services

- **Prime Intellect Compute** — Marketplace aggregating GPU capacity from cloud providers for on-demand AI workloads.
- **PRIME and prime-rl** — Open frameworks for fault-tolerant distributed training and asynchronous reinforcement learning.
- **Environments Hub and verifiers** — Shared task environments and evaluation tools for agent training.
- **Prime Intellect Lab** — Hosted training, evaluation, inference and adapter deployment for specialized agents.
- **Prime Sandboxes** — Linux microVMs for parallel, isolated agent training environments.
- **INTELLECT models** — Open research models demonstrating distributed training and reinforcement learning.

## Achievements

- Trained INTELLECT-1, a 10B model, across five countries and three continents on up to 112 H100 GPUs.
- Released INTELLECT-2, a 32B model trained through globally distributed reinforcement learning.
- Reported more than 10,000 Lab training jobs during its 2026 beta.
- Reported roughly 30 million Sandboxes created before general availability in September 2026.

## Latest updates

- **2026-09** — Prime Sandboxes became generally available; the company reported roughly 30 million sandboxes created.
- **2026-09** — Published a Goodfire case study on detecting reward hacking with shared environments and verifiers.
- **2026-07** — Announced a $130 million Series A led by Radical Ventures.
- **2026-05** — Made Lab generally available after more than 10,000 beta training jobs.

## Links

- Website: https://www.primeintellect.ai/
- LinkedIn: https://www.linkedin.com/company/primeintellect-ai
- Twitter/X: https://x.com/PrimeIntellect
- GitHub: https://github.com/PrimeIntellect-ai

---

Profile page: https://yespress.io/prime-intellect
Published by YesPress — https://yespress.io
Last updated: 2026-09-29