Organization
Prime Intellect
Model developer · Research lab · Compute infrastructure · Developer platform · Dover, Delaware (principal place of business listed in its SEC Form D)
Prime Intellect provides GPU compute, a hosted post-training platform (Lab) with an Environments Hub for reinforcement-learning environments, inference, and code sandboxes, and trains open-weight models such as the INTELLECT series. It maintains the open-source prime-rl training framework and the verifiers environment library.1211
Key facts
- Ownership
- Privately held
- Legal form
- Corporation incorporated in Delaware in 2024, per its SEC Form D filing.10
- Founded
- Unknown
- Sectors
- Open models, Cloud, Research
What is and isn’t public
Prime Intellect publishes INTELLECT model weights on Hugging Face (INTELLECT-3 under the MIT License) and open-sources its prime-rl training framework and verifiers library. INTELLECT-3 is post-trained from a third-party base model (GLM-4.5-Air-Base). Its compute, Lab, and inference offerings are hosted services. This catalog assesses openness per release.1512132
Products
Hosted model products
Labeled Closed product/API — this describes how a documented product is delivered (hosted access). It is not an open artifact and says nothing about other releases by the same organization.
| Product | Kind | Documented delivery and access | Reviewed |
|---|---|---|---|
| Prime Intellect Inference (external site: docs.primeintellect.ai) | Hosted model API | OpenAI-compatible inference API billed per input and output token from a Prime Intellect account balance; API keys are created on the Prime Intellect platform.5 | Sep 29, 2026 |
Other products
| Product | Kind | Documented delivery and access | Reviewed |
|---|---|---|---|
| Prime Intellect Compute (external site: docs.primeintellect.ai) | Compute service | On-demand GPU instances, multi-node clusters, and reserved clusters; the CLI reference documents provisioning instances with the prime CLI and an account API key.123 | Sep 29, 2026 |
| Lab (hosted training and Environments Hub) (external site: docs.primeintellect.ai) | Developer tool | Hosted platform for running reinforcement-learning training on Prime Intellect's infrastructure, launched from the prime CLI with a dashboard for monitoring runs; it includes the Environments Hub and hosted evaluations.42 | Sep 29, 2026 |
Related open artifacts
Model families
- INTELLECTModel family
Model releases
- INTELLECT-3Model releaseAvailability: PublicModel-disclosure tier (USASI rubric v0.1): Open-weight
- INTELLECT-3.1Model releaseAvailability: PublicModel-disclosure tier (USASI rubric v0.1): Open-weight
Research stacks
- prime-rlSoftwareAvailability: Public
Notable documented facts
- INTELLECT-1, a 10B-parameter model trained from scratch on 1 trillion tokens, was trained on up to 14 concurrent nodes across three continents, with compute contributed by 30 independent community contributors, according to its model card.14
- prime-rl is released under Apache 2.0 and verifiers under the MIT License.1213
U.S. eligibility
Eligible · basis: U.S. headquarters
Prime Intellect, Inc. is a Delaware corporation; its July 2026 SEC Form D filing lists its principal place of business in Dover, Delaware, and its terms of service and privacy policy use the same Dover address. Its careers page lists roles in San Francisco and New York City, and its homepage invites applicants to work in San Francisco or remotely. All documented locations are in the United States.1097861
Sources
This listing is not an endorsement or a federal approval.