DataPacts

Prime Intellect: RL Environment Builder Profile

Checked

At a glance

Prime Intellect, Inc. is a US company, incorporated in 2024, that runs the open Environments Hub for reinforcement learning (RL) environments and pays contributors bounties for building environments and evals, alongside selling compute and training to enterprises.

Independent reference. Every buyer is covered on the same terms, every figure carries its source and check date, and no buyer pays to be listed or placed higher.

Request an independent introduction

Tell us what data your company holds. We reply with the buyers that fit and what their terms leave out. Your details reach a buyer only after you agree to a named introduction.

IndependentYour details stay with us until you agreeNot selling is a valid outcome

Contributors send environments and evals to Prime Intellect and get bounties on successful completion; enterprises pay it for compute and training; open Environments Hub, MIT library; rights in a bounty environment not stated.

Its Open Access bounties are "in the $100-500 range", paid on successful completion, according to its program post of October 27, 2025, checked 2026-11-09. It states a USD 130 million Series A in a post of July 8, 2026, while its Form D filed July 1, 2026 reports USD 82,476,632 sold.

Every figure on this page is either Prime Intellect's own statement or a filing, each with its date. The profile sits within AI training data licensing, where an RL environment is one form of data acquired for AI training, and DataPacts has no arrangement with Prime Intellect.

Each value names its source and check date, and how DataPacts sources and dates every figure is set out in the methodology.

What Prime Intellect Builds and Sells

Prime Intellect sells compute and model training to enterprises and runs an open stack around them: the Environments Hub, the verifiers library, the prime-rl trainer, a training platform called Lab, and sandboxes.

The company describes these as one stack, and it states that the stack is open source. In its funding post of July 8, 2026, it claims over 6,000 customers, a figure that covers the whole business rather than the hub alone.

Money moves in two directions. Enterprises pay Prime Intellect for compute and training, and Prime Intellect pays individual contributors bounties for environments, evals and related work.

That makes it both a seller of training and an acquirer of RL environments, and this profile treats it under the second role, because the bounty program is how it acquires material that ends up in AI training.

Within DataPacts' company list, it sits with the platforms, hubs and marketplaces rather than with builders that make environments to order for a single lab.

Some things are not on record as of 2026-11-09. No Prime Intellect program for licensing company data exists in the sources checked, no deal names Prime Intellect, and no lab customer is named.

An RL environment is a task with a starting state and a grader that scores an AI model's attempt, and the page on RL environments covers what they are and who buys them, so this profile stays with Prime Intellect itself.

Prime Intellect Company Identity

Prime Intellect's legal name is Prime Intellect, Inc., a Delaware company registered with the SEC under CIK 2103990, which gave 2024 as its year of incorporation in its Form D. The table lists each attribute with its source, as of 2026-11-09.

Attribute Value Source
Legal name Prime Intellect, Inc. sec.gov (Form D)
Website https://www.primeintellect.ai primeintellect.ai
Registry SEC EDGAR CIK 2103990 (Delaware) sec.gov (Form D)
Country United States sec.gov (Form D)
City Dover, Delaware (issuer address on Form D) sec.gov (Form D)
Incorporated 2024, per Form D sec.gov (Form D)
Type RL environment hub and training platform primeintellect.ai
Officers and directors Vincent Weisser (executive officer and director; signed as Chief Executive Officer); Johannes Hagemann and David Katz (directors) sec.gov (Form D filed 2026-07-01)

The officers and the address come from the Form D filed July 1, 2026, checked 2026-11-10; the Dover address is the issuer address on the filing. Founders are not named in the sources checked, and this page does not take them from profiles, press or encyclopedias.

The Prime Intellect Environments Hub

Prime Intellect launched the Environments Hub on August 27, 2025, as an open place to publish and reuse RL environments built with its verifiers library.

Its library is the one that bounty work must use, and the bounty program described below pays for environments that are published there, so the hub is both a public collection and the destination of paid work.

Growth figures come from Prime Intellect's own posts, each with its date. On October 27, 2025, it stated that more than 400 environments had been crowdsourced in about two months. On July 22, 2026, it counted about 365,000 tasks in 23 tasksets: 198,000 software engineering (SWE) tasks, 28,600 terminal tasks and 137,600 search tasks.

The title of that post reads "365,000+ Environments", but its body counts tasks, not environments, so this page writes "tasks". The difference matters to anyone sizing the hub: a single environment can hold many tasks, and the two numbers of 400 environments and 365,000 tasks measure different things, at dates nine months apart.

No figure on record states how many hub environments came from paid bounties and how many from unpaid contributions, as of 2026-11-09. The SWE share is the largest of the three task types, and it is set beside other builders' coding task sets on the page about RL environments for coding agents.

Prime Intellect's verifiers Library and Its License

verifiers is Prime Intellect's open-source library for building RL environments and evals, and it carries the MIT license, as its GitHub repository showed on November 9, 2026. Bounty work is built with it: the first kind of work the program pays for is environments and evals built with verifiers.

The repository was created on January 22, 2025, and showed 4,776 stars on GitHub as of 2026-11-09, which is a count of users who marked it, not a measure of quality. The MIT license lets anyone use, modify and resell code built on the library, on the condition that the license notice is kept.

It is the license of the library, however, and not a statement about each environment a contributor builds with it: Prime Intellect states no license per contributed environment in the sources checked.

How a grader or rubric scores a model's attempt, and how models learn to game a weak one, is a general topic, covered under verifiers, graders and rubrics.

Prime Intellect Environment Bounties: Terms, Dated

Prime Intellect pays bounties for environment work under its Open-source Environments Program, in two tiers: Open Access tasks that anyone may claim, and Application-Only tasks assigned on approval. The table sets out each term as Prime Intellect states it, checked 2026-11-09.

Term Value Stated by Date Source
Program Open-source Environments Program (Environments Hub bounties) Prime Intellect 2025-10-27 primeintellect.ai/blog/scaling-environments-program
Work wanted RL environments and evals built with verifiers; benchmark reimplementations; paper reproductions; ecosystem integrations Prime Intellect 2025-10-27 primeintellect.ai/blog/scaling-environments-program
Open Access bounty "in the $100-500 range" Prime Intellect 2025-10-27 primeintellect.ai/blog/scaling-environments-program
Application-Only bounty not given on this page n/a n/a n/a
Payment trigger "Bounties are awarded upon successful completion of the task" Prime Intellect 2025-10-27 primeintellect.ai/blog/scaling-environments-program
Grants "hundreds of thousands of $" committed Prime Intellect 2025-10-27 primeintellect.ai/blog/scaling-environments-program
How to claim draft pull request or design doc, or a message to the team; application form for Application-Only tasks Prime Intellect checked 2026-11-09 primeintellect.ai/blog/environments
Eligibility Open Access open to anyone; Application-Only by approval Prime Intellect 2025-10-27 primeintellect.ai/blog/scaling-environments-program
License none stated per environment; "Our entire stack is open-source"; verifiers library MIT Prime Intellect; GitHub checked 2026-11-09 primeintellect.ai/blog/environments; github.com/PrimeIntellect-ai/verifiers
Payment method and tax forms not stated in the sources checked n/a n/a n/a

Every value is Prime Intellect's claim as of the date shown, checked 2026-11-09; the program post can change.

Pay is stated as a range, not a rate. In its post of October 27, 2025, Prime Intellect states that Open Access bounties are "in the $100-500 range" and that "Bounties are awarded upon successful completion of the task", so a claimed task that is not completed earns nothing under those words.

Application-Only tasks are described as larger, but no figure for them is given on this page. The same post states that grants of "hundreds of thousands of $" are committed, without a per-grant amount.

No payment method, currency rule or tax form is stated in the sources checked as of 2026-11-09, so a contributor learns those only from the program itself.

The bounty sits beside every other dated payout and fee in the table of program terms compared.

What Prime Intellect Acquires Through Its Bounties

Prime Intellect's bounties pay for four kinds of work: RL environments and evals built with verifiers, reimplementations of existing benchmarks, reproductions of research papers, and integrations with other tools. The four kinds, as the program post of October 27, 2025 lists them:

  • Environments and evals built with verifiers: new tasks with graders, published to the hub.
  • Benchmark reimplementations: an existing benchmark rebuilt as an environment, complete when the known scores are reproduced.
  • Paper reproductions: a research paper's setup rebuilt so its results can be run again.
  • Ecosystem integrations: connections between the stack and other tools.

A contributor claims an Open Access task by opening a draft pull request or a design doc, or by messaging the team, and applies through a form for an Application-Only task.

The completion test for reimplementations is concrete: the work is done when it reproduces the scores already known for the benchmark, which gives contributor and company a shared check before a bounty is paid.

Rights are the open point. Prime Intellect states no license for each contributed environment and says "Our entire stack is open-source".

Whether a contributor assigns or keeps rights in a bounty environment is not stated in the sources checked as of 2026-11-09. Who owns the code, seed data and tasks inside an environment in general is the subject of rights in an RL environment.

Prime Intellect's Series A Claim Against Its Form D Filings

Prime Intellect states that it raised a USD 130 million Series A led by Radical Ventures, bringing its total funding to over USD 150 million, in a post dated July 8, 2026.

The same post claims over 6,000 customers and over USD 100 million in annualized revenue, all three being Prime Intellect's own figures. The table sets the post beside the two Form D filings in the record, as of 2026-11-09.

Date Document Amount Detail Stated by Source
2026-01-15 Form D USD 49,940,585 sold of USD 64,581,000 offered first sale 2025-12-01; 61 investors filing sec.gov
2026-07-01 Form D USD 82,476,632 sold of USD 84,999,778 offered first sale 2026-06-03; 9 investors; includes USD 30,000,000 of convertible securities filing sec.gov
2026-07-08 company post USD 130,000,000 Series A; over USD 150,000,000 total led by Radical Ventures; over 6,000 customers; over USD 100,000,000 annualized revenue Prime Intellect primeintellect.ai/blog/series-a

Company claims and filings shown side by side; a Form D may cover one closing.

The filings differ from the post. The Form D filed July 1, 2026 reports USD 82,476,632 sold of USD 84,999,778 offered, including USD 30 million of convertible securities, with a first sale on June 3, 2026 and 9 investors.

The earlier Form D, filed January 15, 2026, reports USD 49,940,585 sold of USD 64,581,000 offered, with a first sale on December 1, 2025 and 61 investors.

A Form D can cover a single closing of a round, so the gap between this filing and the figure in the post of July 8, 2026 may reflect later closings; this page shows both and neither reconciles nor sums them.

No valuation is stated here, because none comes from Prime Intellect or a filing. Funding says what investors paid for shares, not what Prime Intellect pays for an environment. Rounds across companies are compared under funding and valuations of AI data companies.

Contributing to Prime Intellect: Benefits, Risks, Downsides and Alternatives

For an environment builder, a Prime Intellect bounty trades a stated, task-sized payment for work that lands in an open hub, so the choice turns on whether the builder wants pay per task or control of the environment.

Benefits

  • A stated Open Access range (see the terms table) and a stated trigger: payment on successful completion.
  • Open Access tasks need no approval.
  • An open-source library under the MIT license, with its star count given above.
  • Grants of "hundreds of thousands of $" stated as committed.

Risks

  • Whether the contributor keeps or assigns rights in a bounty environment is not stated.
  • Payment method, currency and tax handling are not stated.
  • Application-Only pay has no figure on the program page.
  • The amounts are Prime Intellect's claims from a post of October 27, 2025, and can change.

Downsides

  • An environment published in an open hub is public, which makes a later exclusive sale of it to a lab hard to offer.
  • Pay is per task, while other routes on the company list and the price pages show other amounts.
  • No program for licensing company data is on record, so the bounty suits individuals, not a company with a data asset.

Alternatives

  • Selling or licensing the environment directly to labs, or through a marketplace: the guide to how to sell an RL environment or task set compares the routes.
  • Building to order for a custom environment builder.
  • Contributing to an open hub that states no pay, for reputation only.
  • Keeping the environment and not contributing or licensing it at all, which is a valid outcome.

The terms come from a dated post, so the current program page is the reference before relying on any of them.

Prime Intellect Among Open Environment Hubs

Prime Intellect sits with the platforms, hubs and marketplaces on DataPacts' company list, beside BenchFlow, Eigent AI, General Reasoning and HUD. Among them, it is the hub that states bounties for contributions, and the list of RL environment companies sets every company's legal entity and stated funding side by side. Related pages:

  1. Other hubs and marketplaces: General Reasoning, which states no payment to environment creators; BenchFlow, which states no compensation; and HUD, which runs DataVendor, where DataVendor states a 20 % fee on earnings at withdrawal and a USD 1,000 minimum withdrawal, checked 2026-11-09.
  2. Coding builders that build to order: Mechanize, Proximal and Datacurve.
  3. All public price points for RL environments in one place.
  4. The full decision on whether to sell or license data at all, including not selling.

More About Prime Intellect

Prime Intellect Bounties as Contributor Work

A Prime Intellect bounty is paid work on a task, closer to contributor work than to licensing a dataset.

The Open Access range in the terms table is awarded on successful completion, so the contributor is paid for a finished task rather than for granting a license, much as a contributor platform pays per accepted item. Contributor platforms are compared on pay and terms under AI training jobs and platforms.

Frequently Asked Questions About Prime Intellect

What does Prime Intellect do?

Prime Intellect sells compute and model training to enterprises and runs an open stack: the Environments Hub, the verifiers library, the prime-rl trainer, the Lab platform and sandboxes. It also pays contributors bounties for building environments and evals.

How much does Prime Intellect pay?

For bounties, Prime Intellect states Open Access pay "in the $100-500 range", awarded on successful completion (post of 2025-10-27, checked 2026-11-09). No figure for Application-Only tasks is given, and staff pay is not covered here.

Who is the CEO of Prime Intellect?

The Form D filed 2026-07-01 is signed by Vincent Weisser as Chief Executive Officer and lists him as executive officer and director of Prime Intellect, Inc. (checked 2026-11-10). DataPacts does not take roles from social profiles or press, only from a filing or a Prime Intellect page.

Who are Prime Intellect's main competitors?

DataPacts does not rank companies. On its company list, the platforms, hubs and marketplaces segment also includes BenchFlow, Eigent AI, General Reasoning and HUD, each with its own terms; only Prime Intellect among them states bounties.

Is Prime Intellect a real company?

Yes, on the record: Prime Intellect, Inc. is registered with the SEC under CIK 2103990 and filed two Form D notices, on 2026-01-15 and 2026-07-01. Its Form D gives 2024 as its year of incorporation.

Does Prime Intellect buy company data?

No program for licensing company data is on record as of 2026-11-09. Prime Intellect pays bounties to individual contributors for environment work, which is a different relationship from a company data license.