# What is paid to build RL environments and UI gyms > 4 of 12 builders, buyers and platforms for RL environments and UI gyms publish what is paid on the pages we read, read 8 October 2026. AI Sandbox Environments · https://aisandboxenvironments.com/datasets/sandbox-environment-buyers-and-specs/ · computed 2026-10-08 **Payers that publish a figure:** 4 of 12 payers (15 pages read; 23 named). ## The answers, question by question - **Publishes a figure:** no figure published 8, publishes a figure 4 - **Unit of the figure:** not stated 8, a contract or deal 1, per task 1, annual salary 1, per hour of building 1 - **Low end of the stated figure:** not stated 9, a figure in US dollars 2, varies 1 - **High end of the stated figure:** not stated 9, a figure in US dollars 2, varies 1 - **What is built:** stated 11, not stated 1 ## Every payer | Payer | Publishes a figure | Unit of the figure | Low end of the stated figure | High end of the stated figure | What is built | |---|---|---|---|---|---| | Anthropic and OpenAI | publishes a figure | a contract or deal | not stated | not stated | not stated | | Careerflow | publishes a figure | per task | 50 USD | 50 USD | Each task is a small scenario running on a Linux desktop… | | Collinear | no figure published | not stated | not stated | not stated | From GUIs and APIs to MCP, CLI, GitHub, Jira, ServiceNow, and Salesforce | | Deeptune | no figure published | not stated | not stated | not stated | We've built 100s of gyms that mimic popular software like Slack and Salesforce. | | Halluminate | no figure published | not stated | not stated | not stated | It draws on anonymized data from real private transactions, runs inside dynamic desktop… | | HUD | publishes a figure | annual salary | 135000 USD | 230000 USD | HUD's mission is to build reliable, fair and open infrastructure for AI data. | | Keymakr | no figure published | not stated | not stated | not stated | Environments that replicate your target deployment - a Unix shell, a Salesforce CRM… | | Mechanize | no figure published | not stated | not stated | not stated | In these environments, models carry out software engineering work such as building a… | | Mercor | publishes a figure | per hour of building | varies by document | varies by document | Your role will focus on designing, developing, and maintaining high performance RL… | | Prime Intellect | no figure published | not stated | not stated | not stated | DeepSWE long-horizon software engineering tasks from Harbor Hub. | | Surge AI | no figure published | not stated | not stated | not stated | Architect a modular environment framework with clear APIs, curriculum scaffolds, and… | | Veris AI | no figure published | not stated | not stated | not stated | Give coding agents a stateful simulation of your dependencies, so integration tests run… | ## The sentence each answer rests on (75) - Anthropic and OpenAI, [Prolific: 2026 guide to enterprise RL environments](https://www.prolific.com/resources/rl-environments-for-agentic-ai-the-2026-guide-to-building-and-deploying-enterprise-rl-environments), read 2026-10-08, Publishes a figure: "In 2025, Anthropic planned to spend over $1 billion on reinforcement learning environments over the next year, and OpenAI signed multiple seven-figure environment contracts in the same window." - Anthropic and OpenAI, [Prolific: 2026 guide to enterprise RL environments](https://www.prolific.com/resources/rl-environments-for-agentic-ai-the-2026-guide-to-building-and-deploying-enterprise-rl-environments), read 2026-10-08, Unit of the figure: "OpenAI signed multiple seven-figure environment contracts in the same window" - Anthropic and OpenAI, [Prolific: 2026 guide to enterprise RL environments](https://www.prolific.com/resources/rl-environments-for-agentic-ai-the-2026-guide-to-building-and-deploying-enterprise-rl-environments), read 2026-10-08, Low end of the stated figure: No answer on the page, which ran to 35,944 characters when read on 2026-10-08. - Anthropic and OpenAI, [Prolific: 2026 guide to enterprise RL environments](https://www.prolific.com/resources/rl-environments-for-agentic-ai-the-2026-guide-to-building-and-deploying-enterprise-rl-environments), read 2026-10-08, High end of the stated figure: No answer on the page, which ran to 35,944 characters when read on 2026-10-08. - Anthropic and OpenAI, [Prolific: 2026 guide to enterprise RL environments](https://www.prolific.com/resources/rl-environments-for-agentic-ai-the-2026-guide-to-building-and-deploying-enterprise-rl-environments), read 2026-10-08, What is built: No answer on the page, which ran to 35,944 characters when read on 2026-10-08. - Careerflow, [Careerflow: AI/ML Software Engineer, Task Creator (RL Environments)](https://jobs.techstars.com/companies/careerflow/jobs/91150177-ai-ml-software-engineer-task-creator-rl-environments-contract), read 2026-10-08, Publishes a figure: "$50 USD per task, paid once the task is accepted and approved in review." - Careerflow, [Careerflow: AI/ML Software Engineer, Task Creator (RL Environments)](https://jobs.techstars.com/companies/careerflow/jobs/91150177-ai-ml-software-engineer-task-creator-rl-environments-contract), read 2026-10-08, Unit of the figure: "$50 USD per task, paid once the task is accepted and approved in review." - Careerflow, [Careerflow: AI/ML Software Engineer, Task Creator (RL Environments)](https://jobs.techstars.com/companies/careerflow/jobs/91150177-ai-ml-software-engineer-task-creator-rl-environments-contract), read 2026-10-08, Low end of the stated figure: "$50 USD per task, paid once the task is accepted and approved in review." - Careerflow, [Careerflow: AI/ML Software Engineer, Task Creator (RL Environments)](https://jobs.techstars.com/companies/careerflow/jobs/91150177-ai-ml-software-engineer-task-creator-rl-environments-contract), read 2026-10-08, High end of the stated figure: "$50 USD per task, paid once the task is accepted and approved in review." - Careerflow, [Careerflow: AI/ML Software Engineer, Task Creator (RL Environments)](https://jobs.techstars.com/companies/careerflow/jobs/91150177-ai-ml-software-engineer-task-creator-rl-environments-contract), read 2026-10-08, What is built: "Each task is a small scenario running on a Linux desktop — real files, real applications, real documents — plus a program that automatically checks whatever the agent produced." - Collinear, [Collinear home page](https://www.collinear.ai/), read 2026-10-08, Publishes a figure: "Every simulation lab is a self-contained world where your agent operates, complete with the users, tools, data, and tasks it will face in production." - Collinear, [Collinear home page](https://www.collinear.ai/), read 2026-10-08, Unit of the figure: No answer on the page, which ran to 2,686 characters when read on 2026-10-08. - Collinear, [Collinear home page](https://www.collinear.ai/), read 2026-10-08, Low end of the stated figure: No answer on the page, which ran to 2,686 characters when read on 2026-10-08. - Collinear, [Collinear home page](https://www.collinear.ai/), read 2026-10-08, High end of the stated figure: No answer on the page, which ran to 2,686 characters when read on 2026-10-08. - Collinear, [Collinear home page](https://www.collinear.ai/), read 2026-10-08, What is built: "From GUIs and APIs to MCP, CLI, GitHub, Jira, ServiceNow, and Salesforce" - Deeptune, [Deeptune home page](https://www.deeptune.com/), read 2026-10-08, Publishes a figure: "We build training gyms for frontier research labs." - Deeptune, [Deeptune home page](https://www.deeptune.com/), read 2026-10-08, Unit of the figure: No answer on the page, which ran to 738 characters when read on 2026-10-08. - Deeptune, [Deeptune home page](https://www.deeptune.com/), read 2026-10-08, Low end of the stated figure: No answer on the page, which ran to 738 characters when read on 2026-10-08. - Deeptune, [Deeptune home page](https://www.deeptune.com/), read 2026-10-08, High end of the stated figure: No answer on the page, which ran to 738 characters when read on 2026-10-08. - Deeptune, [Deeptune home page](https://www.deeptune.com/), read 2026-10-08, What is built: "We've built 100s of gyms that mimic popular software like Slack and Salesforce." - Halluminate, [Halluminate home page](https://www.halluminate.ai/), read 2026-10-08, Publishes a figure: "Halluminate is a data research lab building benchmarks and RL environments to push the frontier of AI in knowledge work." - Halluminate, [Halluminate home page](https://www.halluminate.ai/), read 2026-10-08, Unit of the figure: No answer on the page, which ran to 1,944 characters when read on 2026-10-08. - Halluminate, [Halluminate home page](https://www.halluminate.ai/), read 2026-10-08, Low end of the stated figure: No answer on the page, which ran to 1,944 characters when read on 2026-10-08. - Halluminate, [Halluminate home page](https://www.halluminate.ai/), read 2026-10-08, High end of the stated figure: No answer on the page, which ran to 1,944 characters when read on 2026-10-08. - Halluminate, [Halluminate home page](https://www.halluminate.ai/), read 2026-10-08, What is built: "It draws on anonymized data from real private transactions, runs inside dynamic desktop environments, and spans trajectories that reach hundreds of steps." - Halluminate, [YC launch: Halluminate, data and sandboxes to train computer use AI](https://www.ycombinator.com/launches/Nme-halluminate-data-and-sandboxes-to-train-computer-use-ai), read 2026-10-08, Publishes a figure: "Halluminate is building realistic sandboxes and datasets to train better computer/browser use AI." - Halluminate, [YC launch: Halluminate, data and sandboxes to train computer use AI](https://www.ycombinator.com/launches/Nme-halluminate-data-and-sandboxes-to-train-computer-use-ai), read 2026-10-08, Unit of the figure: No answer on the page, which ran to 3,950 characters when read on 2026-10-08. - Halluminate, [YC launch: Halluminate, data and sandboxes to train computer use AI](https://www.ycombinator.com/launches/Nme-halluminate-data-and-sandboxes-to-train-computer-use-ai), read 2026-10-08, Low end of the stated figure: No answer on the page, which ran to 3,950 characters when read on 2026-10-08. - Halluminate, [YC launch: Halluminate, data and sandboxes to train computer use AI](https://www.ycombinator.com/launches/Nme-halluminate-data-and-sandboxes-to-train-computer-use-ai), read 2026-10-08, High end of the stated figure: No answer on the page, which ran to 3,950 characters when read on 2026-10-08. - Halluminate, [YC launch: Halluminate, data and sandboxes to train computer use AI](https://www.ycombinator.com/launches/Nme-halluminate-data-and-sandboxes-to-train-computer-use-ai), read 2026-10-08, What is built: "Halluminate is building realistic sandboxes and datasets to train better computer/browser use AI." - HUD, [HUD: Research Engineer (YC jobs)](https://www.ycombinator.com/companies/hud/jobs/9sXeoEj-research-engineer-general), read 2026-10-08, Publishes a figure: "San Francisco (and other major US cities): $135,000 - $230,000" - HUD, [HUD: Research Engineer (YC jobs)](https://www.ycombinator.com/companies/hud/jobs/9sXeoEj-research-engineer-general), read 2026-10-08, Unit of the figure: "Actual offers are adjusted for experience and location, but our base salary bands are" - HUD, [HUD: Research Engineer (YC jobs)](https://www.ycombinator.com/companies/hud/jobs/9sXeoEj-research-engineer-general), read 2026-10-08, Low end of the stated figure: "San Francisco (and other major US cities): $135,000 - $230,000" - HUD, [HUD: Research Engineer (YC jobs)](https://www.ycombinator.com/companies/hud/jobs/9sXeoEj-research-engineer-general), read 2026-10-08, High end of the stated figure: "San Francisco (and other major US cities): $135,000 - $230,000" - HUD, [HUD: Research Engineer (YC jobs)](https://www.ycombinator.com/companies/hud/jobs/9sXeoEj-research-engineer-general), read 2026-10-08, What is built: "HUD's mission is to build reliable, fair and open infrastructure for AI data." - Keymakr, [Keymakr: virtual environments and RL gyms](https://keymakr.com/virtual-environments.html), read 2026-10-08, Publishes a figure: "We can help build, manage, and scale high-fidelity reinforcement learning environments (RL-Gyms) where your agents can learn to reason, plan, and recover from errors." - Keymakr, [Keymakr: virtual environments and RL gyms](https://keymakr.com/virtual-environments.html), read 2026-10-08, Unit of the figure: No answer on the page, which ran to 6,866 characters when read on 2026-10-08. - Keymakr, [Keymakr: virtual environments and RL gyms](https://keymakr.com/virtual-environments.html), read 2026-10-08, Low end of the stated figure: No answer on the page, which ran to 6,866 characters when read on 2026-10-08. - Keymakr, [Keymakr: virtual environments and RL gyms](https://keymakr.com/virtual-environments.html), read 2026-10-08, High end of the stated figure: No answer on the page, which ran to 6,866 characters when read on 2026-10-08. - Keymakr, [Keymakr: virtual environments and RL gyms](https://keymakr.com/virtual-environments.html), read 2026-10-08, What is built: "Environments that replicate your target deployment - a Unix shell, a Salesforce CRM sandbox, or a live e-commerce site replica." - Mechanize, [Mechanize home page](https://www.mechanize.work), read 2026-10-08, Publishes a figure: "We build environments and evals for frontier coding agents." - Mechanize, [Mechanize home page](https://www.mechanize.work), read 2026-10-08, Unit of the figure: No answer on the page, which ran to 1,758 characters when read on 2026-10-08. - Mechanize, [Mechanize home page](https://www.mechanize.work), read 2026-10-08, Low end of the stated figure: No answer on the page, which ran to 1,758 characters when read on 2026-10-08. - Mechanize, [Mechanize home page](https://www.mechanize.work), read 2026-10-08, High end of the stated figure: No answer on the page, which ran to 1,758 characters when read on 2026-10-08. - Mechanize, [Mechanize home page](https://www.mechanize.work), read 2026-10-08, What is built: "In these environments, models carry out software engineering work such as building a feature, deploying an application, or debugging an issue in an unfamiliar codebase." - Mercor, [Mercor: Experienced Software Engineers (RL environments)](https://work.mercor.com/jobs/list_AAABmGFJ15HzTwuYrydOQJn2/experienced-software-engineers), read 2026-10-08, Publishes a figure: "New opportunity Early applicant $75-$100 per hour" - Mercor, [Mercor: Experienced Software Engineers (RL environments)](https://work.mercor.com/jobs/list_AAABmGFJ15HzTwuYrydOQJn2/experienced-software-engineers), read 2026-10-08, Unit of the figure: "New opportunity Early applicant $75-$100 per hour" - Mercor, [Mercor: Experienced Software Engineers (RL environments)](https://work.mercor.com/jobs/list_AAABmGFJ15HzTwuYrydOQJn2/experienced-software-engineers), read 2026-10-08, Low end of the stated figure: "New opportunity Early applicant $75-$100 per hour" - Mercor, [Mercor: Experienced Software Engineers (RL environments)](https://work.mercor.com/jobs/list_AAABmGFJ15HzTwuYrydOQJn2/experienced-software-engineers), read 2026-10-08, High end of the stated figure: "New opportunity Early applicant $75-$100 per hour" - Mercor, [Mercor: Experienced Software Engineers (RL environments)](https://work.mercor.com/jobs/list_AAABmGFJ15HzTwuYrydOQJn2/experienced-software-engineers), read 2026-10-08, What is built: "Your role will focus on designing, developing, and maintaining high performance RL environments." - Mercor, [Mercor: Technical Reviewer, RL Environment Terminal Benchmarking (India)](https://work.mercor.com/jobs/list_AAABmWu3eSKR9nSVhWNLZajY/technical-reviewer-rl-environment-terminal-benchmarking-agentic-ai), read 2026-10-08, Publishes a figure: "1 hired this month Early applicant $25-$30 per hour" - Mercor, [Mercor: Technical Reviewer, RL Environment Terminal Benchmarking (India)](https://work.mercor.com/jobs/list_AAABmWu3eSKR9nSVhWNLZajY/technical-reviewer-rl-environment-terminal-benchmarking-agentic-ai), read 2026-10-08, Unit of the figure: "1 hired this month Early applicant $25-$30 per hour" - Mercor, [Mercor: Technical Reviewer, RL Environment Terminal Benchmarking (India)](https://work.mercor.com/jobs/list_AAABmWu3eSKR9nSVhWNLZajY/technical-reviewer-rl-environment-terminal-benchmarking-agentic-ai), read 2026-10-08, Low end of the stated figure: "1 hired this month Early applicant $25-$30 per hour" - Mercor, [Mercor: Technical Reviewer, RL Environment Terminal Benchmarking (India)](https://work.mercor.com/jobs/list_AAABmWu3eSKR9nSVhWNLZajY/technical-reviewer-rl-environment-terminal-benchmarking-agentic-ai), read 2026-10-08, High end of the stated figure: "1 hired this month Early applicant $25-$30 per hour" - Mercor, [Mercor: Technical Reviewer, RL Environment Terminal Benchmarking (India)](https://work.mercor.com/jobs/list_AAABmWu3eSKR9nSVhWNLZajY/technical-reviewer-rl-environment-terminal-benchmarking-agentic-ai), read 2026-10-08, What is built: "To review, validate, and improve reinforcement learning environment benchmarking pipelines" - Prime Intellect, [Prime Intellect Environments Hub listing](https://app.primeintellect.ai/dashboard/environments), read 2026-10-08, Publishes a figure: No answer on the page, which ran to 7,764 characters when read on 2026-10-08. - Prime Intellect, [Prime Intellect Environments Hub listing](https://app.primeintellect.ai/dashboard/environments), read 2026-10-08, Unit of the figure: No answer on the page, which ran to 7,764 characters when read on 2026-10-08. - Prime Intellect, [Prime Intellect Environments Hub listing](https://app.primeintellect.ai/dashboard/environments), read 2026-10-08, Low end of the stated figure: No answer on the page, which ran to 7,764 characters when read on 2026-10-08. - Prime Intellect, [Prime Intellect Environments Hub listing](https://app.primeintellect.ai/dashboard/environments), read 2026-10-08, High end of the stated figure: No answer on the page, which ran to 7,764 characters when read on 2026-10-08. - Prime Intellect, [Prime Intellect Environments Hub listing](https://app.primeintellect.ai/dashboard/environments), read 2026-10-08, What is built: "DeepSWE long-horizon software engineering tasks from Harbor Hub." - Prime Intellect, [Prime Intellect: Environments Hub launch](https://www.primeintellect.ai/blog/environments), read 2026-10-08, Publishes a figure: "If you are interested in RL environments that don't have a bounty figure listed yet, just ask and we'll figure something out based on the difficulty scale we're using." - Prime Intellect, [Prime Intellect: Environments Hub launch](https://www.primeintellect.ai/blog/environments), read 2026-10-08, Unit of the figure: No answer on the page, which ran to 7,807 characters when read on 2026-10-08. - Prime Intellect, [Prime Intellect: Environments Hub launch](https://www.primeintellect.ai/blog/environments), read 2026-10-08, Low end of the stated figure: No answer on the page, which ran to 7,807 characters when read on 2026-10-08. - Prime Intellect, [Prime Intellect: Environments Hub launch](https://www.primeintellect.ai/blog/environments), read 2026-10-08, High end of the stated figure: No answer on the page, which ran to 7,807 characters when read on 2026-10-08. - Prime Intellect, [Prime Intellect: Environments Hub launch](https://www.primeintellect.ai/blog/environments), read 2026-10-08, What is built: "Prime Sandboxes are fully capable Linux virtual machines built for agentic training, with pricing designed for thousands of concurrent instances." - Surge AI, [Surge AI: RL Environments Architect](https://surgehq.ai/careers/rl-environments-architect), read 2026-10-08, Publishes a figure: "As an RL Environments Architect, you’ll design, instrument, and govern the simulated worlds where agents learn" - Surge AI, [Surge AI: RL Environments Architect](https://surgehq.ai/careers/rl-environments-architect), read 2026-10-08, Unit of the figure: No answer on the page, which ran to 3,568 characters when read on 2026-10-08. - Surge AI, [Surge AI: RL Environments Architect](https://surgehq.ai/careers/rl-environments-architect), read 2026-10-08, Low end of the stated figure: No answer on the page, which ran to 3,568 characters when read on 2026-10-08. - Surge AI, [Surge AI: RL Environments Architect](https://surgehq.ai/careers/rl-environments-architect), read 2026-10-08, High end of the stated figure: No answer on the page, which ran to 3,568 characters when read on 2026-10-08. - Surge AI, [Surge AI: RL Environments Architect](https://surgehq.ai/careers/rl-environments-architect), read 2026-10-08, What is built: "Architect a modular environment framework with clear APIs, curriculum scaffolds, and configurable reward/termination schemas" - Veris AI, [Veris AI home page](https://www.veris.ai/), read 2026-10-08, Publishes a figure: "A simulated copy of your systems, data, APIs, and users, where AI gets tested, trained, and built against before it ever touches production." - Veris AI, [Veris AI home page](https://www.veris.ai/), read 2026-10-08, Unit of the figure: No answer on the page, which ran to 3,011 characters when read on 2026-10-08. - Veris AI, [Veris AI home page](https://www.veris.ai/), read 2026-10-08, Low end of the stated figure: No answer on the page, which ran to 3,011 characters when read on 2026-10-08. - Veris AI, [Veris AI home page](https://www.veris.ai/), read 2026-10-08, High end of the stated figure: No answer on the page, which ran to 3,011 characters when read on 2026-10-08. - Veris AI, [Veris AI home page](https://www.veris.ai/), read 2026-10-08, What is built: "Give coding agents a stateful simulation of your dependencies, so integration tests run and issues get fixed before the PR opens" ## Read and set aside - [Mechanize: announcement](https://www.mechanize.work/announcing-mechanize-inc/): Mechanize's announcement sets out its thesis and the size of the labour market; it states nothing about environments built or what is paid for them - [HUD: platforms for selling a codebase](https://www.hud.ai/resources/best-platforms-selling-codebase): HUD's guide is about selling a codebase as training data, not about building RL environments or what is paid for them - [Epoch AI: FAQ on RL environments](https://epochai.substack.com/p/an-faq-on-reinforcement-learning): the charter admits it for: reported market figures ($200-$2,000 a task; ~$20k for complex tasks) cited as reported, never as a vendor's price - [SemiAnalysis: RL environments and RL for science](https://newsletter.semianalysis.com/p/rl-environments-and-rl-for-science): the charter admits it for: the reported ~$20,000 per UI-gym site, cited as reported - [rl-list.com directory](https://www.rl-list.com/): rl-list.com is a third-party directory of RL environment companies; its figures are funding rounds, not what is paid for environments - [rl-list.com: computer-use environments](https://www.rl-list.com/rl-environments-for-computer-use): rl-list.com's computer-use page is a third-party directory; its figures are funding rounds, not what is paid for environments ## Methodology Each payer's own postings, pricing and buyer pages were read through estate-fetch and stored whole (sha256 on every row). For every page, five questions were answered from its own words: publishes a figure, unit of the figure, low end of the stated figure, high end of the stated figure, what is built. Every stated answer carries the sentence it rests on, checked against the stored text; 'not stated' means the page read does not address the question. A payer with several pages reads 'varies' where they disagree. 12 payers, 15 pages. ## Sources - [Mercor: Experienced Software Engineers (RL environments)](https://work.mercor.com/jobs/list_AAABmGFJ15HzTwuYrydOQJn2/experienced-software-engineers) (read) - [Mercor: Technical Reviewer, RL Environment Terminal Benchmarking (US)](https://work.mercor.com/jobs/list_AAABmWuwiTty4fq67R9Id4lx/technical-reviewer-rl-environment-terminal-benchmarking-agentic-ai) (unread) - [Mercor: Technical Reviewer, RL Environment Terminal Benchmarking (India)](https://work.mercor.com/jobs/list_AAABmWu3eSKR9nSVhWNLZajY/technical-reviewer-rl-environment-terminal-benchmarking-agentic-ai) (read) - [Careerflow (Techstars board): Task Creator, RL Environments](https://jobs.techstars.com/companies/careerflow/jobs/91150177-ai-ml-software-engineer-task-creator-rl-environments-contract) (read) - [Surge AI: RL Environments Architect](https://surgehq.ai/careers/rl-environments-architect) (read) - [Mechanize](https://www.mechanize.work) (read) - [Mechanize: announcement](https://www.mechanize.work/announcing-mechanize-inc/) (not-terms) - [Halluminate](https://www.halluminate.ai/) (read) - [YC launch: Halluminate](https://www.ycombinator.com/launches/Nme-halluminate-data-and-sandboxes-to-train-computer-use-ai) (read) - [HUD](https://www.hud.ai/) (unread) - [HUD: platforms for selling a codebase](https://www.hud.ai/resources/best-platforms-selling-codebase) (not-terms) - [HUD: Research Engineer (YC)](https://www.ycombinator.com/companies/hud/jobs/9sXeoEj-research-engineer-general) (read) - [Prime Intellect Environments Hub](https://app.primeintellect.ai/dashboard/environments) (read) - [Prime Intellect: Environments Hub launch](https://www.primeintellect.ai/blog/environments) (read) - [Collinear](https://www.collinear.ai/) (read) - [Veris](https://www.veris.ai/) (read) - [Deeptune](https://www.deeptune.com/) (read) - [Keymakr: virtual environments and RL gyms](https://keymakr.com/virtual-environments.html) (read) - [Prolific: 2026 guide to enterprise RL environments](https://www.prolific.com/resources/rl-environments-for-agentic-ai-the-2026-guide-to-building-and-deploying-enterprise-rl-environments) (read) - [Epoch AI: FAQ on RL environments](https://epochai.substack.com/p/an-faq-on-reinforcement-learning) (not-terms) - [SemiAnalysis: RL environments and RL for science](https://newsletter.semianalysis.com/p/rl-environments-and-rl-for-science) (not-terms) - [rl-list.com directory](https://www.rl-list.com/) (not-terms) - [rl-list.com: computer-use environments](https://www.rl-list.com/rl-environments-for-computer-use) (not-terms) ## Cite or embed - 4 of 12 builders, buyers and platforms for RL environments and UI gyms publish what is paid on the pages we read, read 8 October 2026. - Publishes a figure, by payer: no figure published 8 of 12, publishes a figure 4 of 12 (12 payers, read 8 October 2026). - Unit of the figure, by payer: not stated 8 of 12, a contract or deal 1 of 12, per hour of building 1 of 12, per task 1 of 12, annual salary 1 of 12 (12 payers, read 8 October 2026). - Low end of the stated figure, by payer: not stated 9 of 12, a figure in US dollars 2 of 12, varies 1 of 12 (12 payers, read 8 October 2026). - High end of the stated figure, by payer: not stated 9 of 12, a figure in US dollars 2 of 12, varies 1 of 12 (12 payers, read 8 October 2026). - What is built, by payer: stated 11 of 12, not stated 1 of 12 (12 payers, read 8 October 2026). Cite as: "What is paid to build RL environments and UI gyms", updated 2026-10-08, https://aisandboxenvironments.com/datasets/sandbox-environment-buyers-and-specs/. CSV: https://aisandboxenvironments.com/datasets/sandbox-environment-buyers-and-specs.csv JSON: https://aisandboxenvironments.com/datasets/sandbox-environment-buyers-and-specs.json Embed (plain HTML, static link back): see https://aisandboxenvironments.com/datasets/sandbox-environment-buyers-and-specs/#lead