RL environment vendors
Fourteen companies say on their own pages, or in a dated primary source, that they build RL environments or the tools to share and run them. None publishes a price. They split into environment specialists, human data providers that have added environments, and open hubs and training tools.
Last verified 7 November 2026
Environment specialists
| Company | What it says it builds | Delivery | Published detail |
|---|---|---|---|
| Bespoke Labs | RL environments plus the infrastructure to run them: an environment engine, sandboxing and execution, agent optimization. Sells to frontier labs, newer labs and enterprises. | Mixed (built environments and software) | Announced $40M across a seed round and a Series A on 6 Jul 2026. Source |
| Halluminate | Hosted practice copies of business software (its examples: Salesforce, Slack, ticketing systems) where agents that drive a computer or browser are trained and tested. | Managed service | Y Combinator Summer 2025 company; its YC listing was updated 24 Oct 2026. Source |
| Patronus AI | Simulated copies of websites and company systems, which it calls digital world models, used to train agents and to stress-test them; also automated agent evaluation. | Mixed | Not published |
Note on fees: when an introduction to a company in these tables leads to a contract, that company may owe us a referral fee. That never moves its row or alters its description. RL Environment Companies is independent, with no affiliation with, or endorsement from, the firms it names.
Data providers that also build environments
| Company | What it says it builds | Delivery | Published detail |
|---|---|---|---|
| Scale AI | Enterprise tool-use environments and computer-use environments, alongside its wider data and evaluation work. | Mixed | Its RL environments page lists Enterprise Atlas (4 enterprise environments, 2,000+ tasks) and computer-use sets: Professional Office (1,000 tasks) and OSWorld Advanced (370 tasks). Source |
| Turing | Copies of the Jira, Salesforce and Zendesk interfaces for agents to work in, plus backend environments reached over MCP, within its post-training and evaluation business. | Managed service | Not published |
| Surge AI | RL environments, rubrics and verifiers, alongside RLHF, demonstrations and human evaluation for AI labs. | Managed service | Its products page says it works in more than 70 languages. Source |
| Mercor | Expert-built training data, human evaluation, the APEX benchmarks and RL environments for AI labs. | Managed service | Its blog post of 9 Jul 2026 says Mercor is buying Deeptune, whose environments recreate hundreds of enterprise apps, spreadsheets to Salesforce. Source |
| micro1 | Post-training and RL data produced by human experts it recruits and manages. | Managed service | Not published |
| Handshake AI | Its fellowship pays students and subject experts, recruited through the Handshake jobs network, to work on projects that train and evaluate AI lab models. | Managed service | Named by Epoch AI (Jan 2026) among providers selling RL environments. Source |
| Toloka | Auto-verifiable tasks for evaluation and RL, plus demonstrations, preference data and red teaming. | Managed service | Headquartered in Amsterdam. Source |
| Appen | Ready-made RL tasks that come with automatic checks, sold beside its annotation and post-training data. | Managed service | A blog post (Feb 2026, revised Oct 2026) describes a set of 500 coding tasks, every one with its own automatic verifier. Source |
| Centific | RL environments as one part of its AI Data Foundry platform and services, with preference data, fine-tuning data and safety evaluation. | Mixed | Not published |
Open hubs and training tools
| Company | What it says it builds | Delivery | Published detail |
|---|---|---|---|
| Prime Intellect | The Environments Hub, where anyone can publish or reuse RL and evaluation environments; also the prime-rl training code (open source) and a paid hosted trainer. | Mixed (open hub and hosted software) | Launched the Environments Hub on 27 Aug 2025 and runs environment bounties (amounts not published in the launch post). Source |
| OpenPipe | Software that uses reinforcement learning to train agents, and the open-source ART library. Tools rather than built-to-order environments. | Software platform | Now owned by CoreWeave, which announced the purchase on 3 Sept 2025. Source |
Which kind of vendor fits which project?
A specialist suits a team that wants environments engineered around a product or a hard skill, with sandboxing and graders built in. A data provider suits work where the bottleneck is expert time: writing tasks, rubrics and reference solutions in a domain like law or accounting. Open hubs suit a team that wants to start from existing environments and adapt them in-house.
Many projects mix the three: an open environment as the base, a specialist to harden the sandbox, and experts to write tasks. Ask each vendor which parts it does itself and which it subcontracts.
What changed in these companies' published terms?
Dated changes in what the companies above publish, newest first.
Opening check: we read each listed company's pages. Not one of the fourteen prints a price for environments or tasks.
Halluminate's Y Combinator listing updated.
Deeptune, an environment specialist, is being bought by Mercor (Mercor blog). From now on we file Deeptune under Mercor.
Bespoke Labs announced $40M across its seed round and Series A.
Epoch AI published its FAQ on RL environments, the main public source on per-task cost and exclusivity.
CoreWeave announced it would acquire OpenPipe.
Prime Intellect launched the Environments Hub.
Where these facts come from (14)
- Bespoke Labs: Bespoke Labs blog, funding announcement (6 Jul 2026) (page checked 7 November 2026).
- Halluminate: Y Combinator launch page for Halluminate (page checked 7 November 2026).
- Patronus AI: Patronus AI website (page checked 7 November 2026).
- Scale AI: Scale AI RL environments page (page checked 7 November 2026).
- Turing: Turing RL environments page (page checked 7 November 2026).
- Surge AI: Surge AI products page (page checked 7 November 2026).
- Mercor: Mercor blog, Mercor to acquire Deeptune (9 Jul 2026) (page checked 7 November 2026).
- micro1: micro1 website (page checked 7 November 2026).
- Handshake AI: Epoch AI's January 2026 RL environments FAQ, which names it as a seller (page checked 7 November 2026).
- Toloka: Toloka about page (page checked 7 November 2026).
- Appen: Appen blog on RL with verifiable rewards (page checked 7 November 2026).
- Centific: Centific AI Data Foundry page (page checked 7 November 2026).
- Prime Intellect: Prime Intellect blog, Environments Hub launch (27 Aug 2025) (page checked 7 November 2026).
- OpenPipe: CoreWeave news, acquisition of OpenPipe (3 Sept 2025) (page checked 7 November 2026).
Questions about this list
- Why are some well-known data companies missing?
- We list a company only when its own pages, or a dated primary source, show that it builds RL environments today. Where we could not confirm that, or a firm's status was unclear, it is left out until we can.
- Why does every price say "not published"?
- Because none of these companies publishes one for environments or tasks (checked 7 November 2026). The only public cost figures come from Epoch AI's interviews; see task costs.
- Has anyone here tested these builders?
- No. Nothing on this page comes from testing, auditing or ranking. The list is built from published information only, and the order is by kind of company, not by quality.
- Is Deeptune still a separate company?
- Mercor announced on 9 Jul 2026 that it is acquiring Deeptune, so we list Deeptune's work under Mercor.
- How often is this page checked?
- Each company page is re-read when we update the list, and anything that changed goes in the changelog below with its date.
Related pages
- RL environment companies: the main entry
- Task costs: what one task costs, from published figures
- Open-source environments and hubs: what is free to start from
Get a quote
Tell us the company ZIP code, a contact, and an outline of the environments you have in mind. We compare it with what each builder publishes and send you the shortlist by name.
How the site earns: we make introductions, and a builder you later contract with may pay us for yours.
Received. The next three things:
- Your outline is set against what each RL environment builder says it makes and the terms it publishes.
- A shortlist arrives by email: names, links, and a sentence on the fit for each.
- No builder hears about you until you answer yes. There is no mailing list to join.