{
  "name": "RL environment companies dataset, November 2026",
  "publisher": "Patronus AI",
  "url": "https://www.patronus.ai/resources/rl-environment-companies-dataset",
  "terms": "Use is covered by the Patronus AI Terms of Use, https://www.patronus.ai/terms-of-use",
  "as_of": "2026-11-11",
  "note": "A dated snapshot. Each value comes from the source in its *_source field, which is the company’s own pages, announcements, GitHub or papers, or for some founding years and headquarters its LinkedIn page, Y Combinator profile or PR Newswire release. Empty fields are not stated by the company. Companies are alphabetical, not ranked.",
  "companies": [
    {
      "name": "Andon Labs",
      "url": "https://andonlabs.com",
      "group": "Research labs that build environments",
      "founded": {
        "v": "2023",
        "src": "https://www.ycombinator.com/companies/andon-labs"
      },
      "hq": {
        "v": "San Francisco, California",
        "src": "https://andonlabs.com/join",
        "label": "HQ"
      },
      "focus": {
        "v": "Andon Labs studies frontier AI safety by running AI-run businesses in the real world and comparing models in simulated long-horizon evals.",
        "src": "https://andonlabs.com"
      },
      "public_work": [
        {
          "name": "Vending-Bench 2",
          "url": "https://andonlabs.com/evals/vending-bench-2",
          "date": null
        },
        {
          "name": "Drone-Bench",
          "url": "https://andonlabs.com/evals/drone-bench",
          "date": "2026-07-23",
          "dateSrc": "https://andonlabs.com/publications"
        },
        {
          "name": "Butter-Bench",
          "url": "https://arxiv.org/abs/2510.21860",
          "date": "2025-10-23",
          "note": "arXiv v1 date. Andon’s publications page lists Oct 28, 2025"
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "Bespoke Labs",
      "url": "https://bespokelabs.ai",
      "group": "Research labs that build environments",
      "founded": {
        "v": null,
        "src": null
      },
      "hq": {
        "v": "Mountain View, California",
        "src": "https://bespokelabs.ai/about",
        "label": "HQ"
      },
      "focus": {
        "v": "Bespoke Labs builds RL environments and infrastructure so frontier labs and enterprises can train long-horizon agents for production.",
        "src": "https://bespokelabs.ai"
      },
      "public_work": [
        {
          "name": "OpenThoughts",
          "url": "https://github.com/open-thoughts/open-thoughts",
          "date": null
        },
        {
          "name": "Curator",
          "url": "https://github.com/bespokelabsai/curator",
          "date": null
        },
        {
          "name": "AutoResearchExam",
          "url": "https://bespokelabs.ai/blog/introducing-autoresearchexam",
          "date": "2026-09-10"
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "Browserbase",
      "url": "https://www.browserbase.com",
      "group": "Agent infrastructure",
      "founded": {
        "v": "2024",
        "src": "https://www.browserbase.com/about"
      },
      "hq": {
        "v": "San Francisco, California",
        "src": "https://www.browserbase.com/terms-of-service",
        "label": "HQ"
      },
      "focus": {
        "v": "Browserbase runs cloud browsers for AI agents, and its BrowserEnv wraps those sessions as an environment to train and evaluate browser agents.",
        "src": "https://www.browserbase.com/browserenv"
      },
      "public_work": [
        {
          "name": "Stagehand",
          "url": "https://www.browserbase.com/stagehand",
          "date": null
        },
        {
          "name": "BrowserEnv",
          "url": "https://www.browserbase.com/browserenv",
          "date": null
        },
        {
          "name": "Browserbase MCP server",
          "url": "https://github.com/browserbase/mcp-server-browserbase",
          "date": null
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "Chakra Labs",
      "url": "https://www.chakra.dev",
      "group": "Software replicas and simulated workplaces",
      "founded": {
        "v": null,
        "src": null
      },
      "hq": {
        "v": "New York, New York",
        "src": "https://www.chakra.dev/journal/terms-of-service",
        "label": "offices"
      },
      "focus": {
        "v": "Chakra Labs is a research and design lab building expert-authored tasks and pixel-perfect environments to train models and evaluate real-world capability.",
        "src": "https://www.chakra.dev"
      },
      "public_work": [
        {
          "name": "MAGI",
          "url": "https://www.chakra.dev/magi-bench",
          "date": "2026-07-20",
          "dateSrc": "https://www.chakra.dev/journal/magi-benchmark"
        },
        {
          "name": "GDPx legal domain tasks (SemiAnalysis)",
          "url": "https://www.chakra.dev/journal/semianalysis-gdpx-legal",
          "date": "2026-11-01",
          "dateSrc": "https://x.com/chakra_ai/status/2117080458555257018"
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "E2B",
      "url": "https://e2b.dev",
      "group": "Agent infrastructure",
      "founded": {
        "v": "2023",
        "src": "https://e2b.dev/about"
      },
      "hq": {
        "v": "San Francisco, California and Prague, Czech Republic",
        "src": "https://e2b.dev/careers",
        "label": "offices"
      },
      "focus": {
        "v": "E2B runs isolated microVM sandboxes for AI agents, in its cloud or the customer's, for coding, computer use, background jobs and RL rollouts.",
        "src": "https://e2b.dev"
      },
      "public_work": [
        {
          "name": "E2B",
          "url": "https://github.com/e2b-dev/E2B",
          "date": null
        },
        {
          "name": "E2B Desktop",
          "url": "https://github.com/e2b-dev/desktop",
          "date": null
        },
        {
          "name": "E2B runtime",
          "url": "https://github.com/e2b-dev/runtime",
          "date": null
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "Fleet",
      "url": "https://www.fleetai.com",
      "group": "Software replicas and simulated workplaces",
      "founded": {
        "v": null,
        "src": null
      },
      "hq": {
        "v": "San Francisco, California",
        "src": "https://www.fleetai.com/careers",
        "label": "HQ",
        "note": "Its LinkedIn lists New York"
      },
      "focus": {
        "v": "Fleet builds training gyms for agents, high-fidelity simulation environments where AI practices tasks and humans provide supervision.",
        "src": "https://www.fleetai.com/about"
      },
      "public_work": [
        {
          "name": "WitnessBench",
          "url": "https://arxiv.org/abs/2609.32208",
          "date": "2026-09-26"
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "General Reasoning",
      "url": "https://www.gr.inc",
      "group": "Research labs that build environments",
      "founded": {
        "v": "2025",
        "src": "https://www.linkedin.com/company/general-reasoning"
      },
      "hq": {
        "v": "London, United Kingdom",
        "src": "https://www.linkedin.com/company/general-reasoning",
        "label": "HQ"
      },
      "focus": {
        "v": "An AI research company training models for long-horizon work, with environment design as a research area and OpenReward, an open platform serving RL environments.",
        "src": "https://www.gr.inc"
      },
      "public_work": [
        {
          "name": "OpenReward",
          "url": "https://www.gr.inc/openreward",
          "date": null
        },
        {
          "name": "Open Reward Standard (ORS)",
          "url": "https://www.gr.inc/releases/the-open-reward-standard",
          "date": null
        },
        {
          "name": "KellyBench",
          "url": "https://www.gr.inc/releases/introducing-kellybench",
          "date": "2026-04-09"
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "Good Start Labs",
      "url": "https://goodstartlabs.com",
      "group": "Game environments",
      "founded": {
        "v": "2025",
        "src": "https://www.linkedin.com/company/good-start-labs"
      },
      "hq": {
        "v": null,
        "src": null,
        "label": null
      },
      "focus": {
        "v": "Builds reinforcement learning environments and world-model data from games, plus public benchmarks that measure model capability in games.",
        "src": "https://goodstartlabs.com/about"
      },
      "public_work": [
        {
          "name": "AI Diplomacy",
          "url": "https://github.com/GoodStartLabs/AI_Diplomacy",
          "date": "2025-06-05",
          "dateSrc": "https://goodstartlabs.com/research"
        },
        {
          "name": "Democratizing Diplomacy: A Harness for Evaluating Any Large Language Model on Full-Press Diplomacy",
          "url": "https://arxiv.org/abs/2508.07485",
          "date": "2025-08-10"
        },
        {
          "name": "Humor leaderboard (Bad Cards)",
          "url": "https://goodstartlabs.com/benchmarks",
          "date": null
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "Halluminate",
      "url": "https://halluminate.ai",
      "group": "Domain specialists",
      "founded": {
        "v": "2024",
        "src": "https://www.ycombinator.com/companies/halluminate"
      },
      "hq": {
        "v": "San Francisco",
        "src": "https://www.ycombinator.com/companies/halluminate",
        "label": "HQ"
      },
      "focus": {
        "v": "A data research lab building benchmarks and RL environments for knowledge work, starting with financial services and consulting.",
        "src": "https://halluminate.ai"
      },
      "public_work": [
        {
          "name": "Westworld Finance Diligence Bench",
          "url": "https://halluminate.ai/blog/due-diligence-bench",
          "date": "2026-08-11"
        },
        {
          "name": "Westworld v1",
          "url": "https://halluminate.ai/blog/westworld",
          "date": "2025-11-23"
        },
        {
          "name": "Web Bench",
          "url": "https://halluminate.ai/blog/benchmark",
          "date": "2025-05-30"
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "HUD",
      "url": "https://www.hud.ai",
      "group": "Open hubs and platforms",
      "founded": {
        "v": "2025",
        "src": "https://www.ycombinator.com/companies/hud"
      },
      "hq": {
        "v": "San Francisco, California",
        "src": "https://www.linkedin.com/company/hud-evals",
        "label": "HQ"
      },
      "focus": {
        "v": "A platform and SDK for building RL environments, turning expertise into environments that train and evaluate models, with QA agents auditing traces.",
        "src": "https://www.hud.ai"
      },
      "public_work": [
        {
          "name": "hud-python SDK",
          "url": "https://github.com/hud-evals/hud-python",
          "date": null
        },
        {
          "name": "SheetBench-50",
          "url": "https://www.hud.ai/research/sheetbench-50",
          "date": "2025-10-01"
        },
        {
          "name": "Assemble Bench",
          "url": "https://www.hud.ai/research/assemble-benchmark",
          "date": "2026-07-31"
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "Idler",
      "url": "https://idler.ai",
      "group": "Software replicas and simulated workplaces",
      "founded": {
        "v": "2025",
        "src": "https://www.ycombinator.com/companies/idler"
      },
      "hq": {
        "v": "San Francisco",
        "src": "https://www.ycombinator.com/companies/idler",
        "label": "HQ"
      },
      "focus": {
        "v": "A frontier data lab building RL environments from real production work, across frontier coding, economically valuable workflows and bespoke evaluations.",
        "src": "https://idler.ai",
        "src2": "https://idler.ai/careers"
      },
      "public_work": [
        {
          "name": "ShelfLife",
          "url": "https://idler.ai/collections/shelflife",
          "date": null
        },
        {
          "name": "ShelfLife E-Sim",
          "url": "https://idler.ai/collections/shelflife-e-sim",
          "date": null
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "Invisible Technologies",
      "url": "https://invisibletech.ai",
      "group": "Expert data and evaluation companies",
      "founded": {
        "v": "2015",
        "src": "https://invisibletech.ai/about"
      },
      "hq": {
        "v": null,
        "src": null,
        "label": null
      },
      "focus": {
        "v": "Builds expert-built RL environments where domain experts define task logic, design reward rubrics and annotate trajectories, alongside expert data.",
        "src": "https://invisibletech.ai/solutions/rl-environments"
      },
      "public_work": [],
      "as_of": "2026-11-11"
    },
    {
      "name": "Labelbox",
      "url": "https://labelbox.com",
      "group": "Expert data and evaluation companies",
      "founded": {
        "v": "2018",
        "src": "https://labelbox.com/company/about"
      },
      "hq": {
        "v": "San Francisco, California",
        "src": "https://labelbox.com/company/about",
        "label": "HQ"
      },
      "focus": {
        "v": "Labelbox builds RL training environments and evals for post-training on reasoning, tool use and computer use, co-designed with domain experts.",
        "src": "https://labelbox.com/products/rl/"
      },
      "public_work": [
        {
          "name": "Implicit Intelligence",
          "url": "https://labelbox.com/blog/implicit-intelligence-and-agent-as-a-world-evaluating-agents-on-what-users-dont-say/",
          "date": "2025-11-21"
        },
        {
          "name": "EchoChain",
          "url": "https://labelbox.com/blog/introducing-echochain-an-audio-benchmark-for-reasoning-under-pressure-in-full-duplex-dialogue/",
          "date": "2026-03-04"
        },
        {
          "name": "Labelbox Leaderboards",
          "url": "https://labelbox.com/leaderboards/",
          "date": null
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "Mechanize",
      "url": "https://www.mechanize.work/",
      "group": "Domain specialists",
      "founded": {
        "v": null,
        "src": null
      },
      "hq": {
        "v": null,
        "src": null,
        "label": null
      },
      "focus": {
        "v": "Mechanize builds RL environments and evals in which frontier coding agents do software engineering tasks scored by a grader.",
        "src": "https://www.mechanize.work/"
      },
      "public_work": [
        {
          "name": "GBA Eval",
          "url": "https://www.mechanize.work/technical-blog/introducing-gba-eval/",
          "date": "2026-05-02"
        },
        {
          "name": "gba-eval (GitHub)",
          "url": "https://github.com/mechanize-work/gba-eval",
          "date": null
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "Mercor",
      "url": "https://www.mercor.com",
      "group": "Expert data and evaluation companies",
      "founded": {
        "v": "2023",
        "src": "https://mercor.com"
      },
      "hq": {
        "v": "San Francisco, California",
        "src": "https://mercor.com",
        "label": "HQ"
      },
      "focus": {
        "v": "Mercor supplies expert human work for frontier research, AI benchmarks and AI agent training for AI labs and enterprises.",
        "src": "https://mercor.com"
      },
      "public_work": [
        {
          "name": "APEX-Agents",
          "url": "https://arxiv.org/abs/2601.14242",
          "date": "2026-01-20"
        },
        {
          "name": "APEX-SWE",
          "url": "https://www.mercor.com/blog/introducing-apex-swe/",
          "date": "2026-03-24"
        },
        {
          "name": "APEX-Accounting",
          "url": "https://www.mercor.com/blog/introducing-the-ai-productivity-index-for-accounting/",
          "date": "2026-07-31"
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "Patronus AI",
      "url": "https://www.patronus.ai",
      "group": "Research labs that build environments",
      "founded": {
        "v": "2023",
        "src": "https://www.patronus.ai/company"
      },
      "hq": {
        "v": "San Francisco",
        "src": "https://www.patronus.ai/join",
        "label": "HQ"
      },
      "focus": {
        "v": "Patronus AI is a behavioral research lab for AI that studies where models fail and builds simulated environments and training data to improve them.",
        "src": "https://www.patronus.ai/"
      },
      "public_work": [
        {
          "name": "GlobeBench",
          "url": "https://www.patronus.ai/research/globebench",
          "date": "2026-10-15"
        },
        {
          "name": "TRAIL",
          "url": "https://www.patronus.ai/research/trail",
          "date": "2025-05-13"
        },
        {
          "name": "FinanceBench",
          "url": "https://www.patronus.ai/research/financebench",
          "date": "2023-11-20"
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "Plato",
      "url": "https://plato.so/",
      "group": "Software replicas and simulated workplaces",
      "founded": {
        "v": "2025",
        "src": "https://www.prnewswire.com/news-releases/plato-raises-16-million-to-build-the-training-grounds-for-ais-next-leap-into-real-work-302921832.html"
      },
      "hq": {
        "v": "San Francisco, California",
        "src": "https://plato.so/landing/",
        "label": "HQ"
      },
      "focus": {
        "v": "Plato builds simulated worlds of the software and tasks of real jobs, where models train and are evaluated on that work.",
        "src": "https://plato.so/landing/"
      },
      "public_work": [
        {
          "name": "Computer Work Val v1.0",
          "url": "https://plato.so/research/computer-work-val/",
          "date": "2026-10-27"
        },
        {
          "name": "Web Clone Bench v1",
          "url": "https://plato.so/research/web-clone-bench/",
          "date": "2026-10-05"
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "Prime Intellect",
      "url": "https://www.primeintellect.ai/",
      "group": "Open hubs and platforms",
      "founded": {
        "v": null,
        "src": null
      },
      "hq": {
        "v": null,
        "src": null,
        "label": null
      },
      "focus": {
        "v": "Prime Intellect offers an open-source library and hub for building and sharing RL environments, plus hosted RL training, evals and sandboxes.",
        "src": "https://www.primeintellect.ai/"
      },
      "public_work": [
        {
          "name": "Environments Hub",
          "url": "https://www.primeintellect.ai/blog/environments",
          "date": "2025-08-27"
        },
        {
          "name": "Verifiers",
          "url": "https://github.com/PrimeIntellect-ai/verifiers",
          "date": null
        },
        {
          "name": "INTELLECT-3",
          "url": "https://www.primeintellect.ai/blog/intellect-3",
          "date": "2025-11-26"
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "Scale AI",
      "url": "https://scale.com",
      "group": "Expert data and evaluation companies",
      "founded": {
        "v": "2016",
        "src": "https://scale.com/about"
      },
      "hq": {
        "v": "San Francisco, California",
        "src": "https://scale.com/about",
        "label": "HQ"
      },
      "focus": {
        "v": "Scale AI's Enterprise Atlas RL environments train agents on long-horizon work across real enterprise systems and desktop environments.",
        "src": "https://scale.com/rlenvironments"
      },
      "public_work": [
        {
          "name": "SWE Atlas",
          "url": "https://scale.com/blog/swe-atlas",
          "date": "2026-03-04"
        },
        {
          "name": "AgentEnv",
          "url": "https://labs.scale.com/blog/introducing-agentenv",
          "date": "2026-10-05"
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "Snorkel AI",
      "url": "https://snorkel.ai",
      "group": "Expert data and evaluation companies",
      "founded": {
        "v": "2019",
        "src": "https://snorkel.ai/company/"
      },
      "hq": {
        "v": "San Francisco, California; New York, New York",
        "src": "https://snorkel.ai/contact/",
        "label": "offices"
      },
      "focus": {
        "v": "Snorkel AI develops specialized training data and environments for frontier labs and AI teams' models and agents.",
        "src": "https://snorkel.ai"
      },
      "public_work": [
        {
          "name": "Agentic Coding 2.0",
          "url": "https://snorkel.ai/leaderboard/agentic-coding-2/",
          "date": null
        },
        {
          "name": "SnorkelFinance 2.0",
          "url": "https://snorkel.ai/leaderboard/snorkel-finance-2-0/",
          "date": null
        },
        {
          "name": "SnorkelLegal",
          "url": "https://snorkel.ai/leaderboard/snorkel-legal/",
          "date": null
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "Surge AI",
      "url": "https://surgehq.ai",
      "group": "Expert data and evaluation companies",
      "founded": {
        "v": "2020",
        "src": "https://www.linkedin.com/company/surge-ai"
      },
      "hq": {
        "v": "San Francisco, California",
        "src": "https://www.linkedin.com/company/surge-ai",
        "label": "HQ"
      },
      "focus": {
        "v": "Surge AI sells off-the-shelf training data, evals and RL environments built by the experts behind its frontier-lab work.",
        "src": "https://surgehq.ai/ots"
      },
      "public_work": [
        {
          "name": "DAYJOB: Tax",
          "url": "https://surgehq.ai/benchmarks/dayjob-tax",
          "date": null
        },
        {
          "name": "DAYJOB: Finance",
          "url": "https://surgehq.ai/benchmarks/dayjob-finance",
          "date": null
        },
        {
          "name": "RL Environments and the Hierarchy of Agentic Capabilities",
          "url": "https://surgehq.ai/blog/rl-envs-real-world",
          "date": "2025-11-03"
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "Turing",
      "url": "https://www.turing.com",
      "group": "Expert data and evaluation companies",
      "founded": {
        "v": "2018",
        "src": "https://help.turing.com/articles/2834113361-the-turing-story"
      },
      "hq": {
        "v": "San Francisco, California",
        "src": "https://www.turing.com/company",
        "label": "HQ"
      },
      "focus": {
        "v": "Turing builds iterable UI and non-UI RL environments, including app replicas and MCP backends, for agents to learn long-horizon workflows.",
        "src": "https://www.turing.com/frontier-ai/rl-environments"
      },
      "public_work": [
        {
          "name": "Turing Terminal-Bench Data",
          "url": "https://labs.turing.com/training-sets/terminal-bench",
          "date": null
        },
        {
          "name": "Turing Frontier Agentic STEM",
          "url": "https://labs.turing.com/training-sets/frontier-agentic-stem",
          "date": null
        }
      ],
      "as_of": "2026-11-11"
    },
    {
      "name": "Veris AI",
      "url": "https://veris.ai",
      "group": "Software replicas and simulated workplaces",
      "founded": {
        "v": null,
        "src": null
      },
      "hq": {
        "v": "New York, New York",
        "src": "https://www.linkedin.com/company/veris-ai/about/",
        "label": "HQ"
      },
      "focus": {
        "v": "Veris AI builds a simulated copy of a company's systems, data, APIs and users where AI agents are tested and trained.",
        "src": "https://veris.ai"
      },
      "public_work": [
        {
          "name": "VAmoS Bench",
          "url": "https://veris.ai/leaderboard/vamos",
          "date": null
        },
        {
          "name": "VAmoS Pro Bench",
          "url": "https://veris.ai/blog/vamos-pro-bench-a-harder-exam-for-voice-agents",
          "date": "2026-10-01"
        },
        {
          "name": "Veris FinAgent Arena",
          "url": "https://veris.ai/leaderboard/arena",
          "date": null
        }
      ],
      "as_of": "2026-11-11"
    }
  ]
}