Mercor vs Scale AI: How They Compare in 2026
Patronus AI TeamNovember 20267 min read
Patronus AI, which builds RL environments, publishes this comparison and is neither company. Every fact about each company comes from its own pages, read on Nov 11, 2026 and linked under its section.
As of November 2026
Mercor and Scale AI both supply expert human data, evaluations and RL environments to AI labs. Mercor, founded in 2023, is built around a network of domain experts and its APEX benchmarks of professional work. Scale AI, founded in 2016, pairs its data business with AI applications for enterprises and governments and lists FedRAMP High. Choose Mercor for professional-domain experts and benchmarks, and Scale AI for one vendor across data, evaluation and government-grade security.
Both companies sell human data and, as of November 2026, also build RL environments, so the differences are in where each came from, who does the work and what else each sells. This page compares them as of November 2026, from each company’s own pages.
Snapshot as of
Mercor and Scale AI at a glance
Each cell is what the company states on its own pages, as of Nov 11, 2026.
| Criterion | Mercor | Scale AI |
|---|---|---|
| In its own words | “A leading AI data company” | “Reliable AI for the world’s most important decisions” |
| Founded, headquarters | 2023, San Francisco | 2016, San Francisco |
| Leadership | Co-founders and co-CEOs Brendan Foody and Adarsh Hiremath | CEO Francis deSouza (since Aug 10, 2026); founder Alexandr Wang is chairman |
| What it sells | Expert human data, evaluations, benchmarks and RL environments, plus enterprise AI deployment | Training data, evaluations, RL environments and AI applications for enterprises and governments |
| Who does the work | A network of domain experts vetted with AI interviews | Subject matter experts, through its Outlier platform |
| RL environments | Built in three steps (world, tools, tasks and verifiers); Deeptune acquisition announced Jul 9, 2026 | Scale RL Environments, introduced Feb 27, 2026, sold as Enterprise Atlas as of Nov 11, 2026, on MCP servers, CLI and GUIs |
| Benchmarks and research | APEX benchmarks, and hosted partner leaderboards such as Harvey LAB | Scale Labs research and public leaderboards |
| Security (its own site) | SOC 2 Type 2 | SOC 2 Type II, ISO 27001, FedRAMP High, DoD IL4 |
| How to buy | Contact its team or book a demo | Book a demo or request samples |
Mercor
Mercor calls itself “a leading AI data company, building the layer between human expertise and frontier models.” It was founded in 2023 and is headquartered in San Francisco, with offices in New York and London. Its trust center describes Mercor as “a better way to hire,” using AI to match talent with opportunities.
Its co-founders and co-CEOs are Brendan Foody and Adarsh Hiremath, as of Nov 11, 2026.
What Mercor offers
- Human data and evaluations from domain experts, such as physicians, lawyers, engineers and consultants, vetted with role-specific AI interviews.
- RL environments, which it builds in three steps: a realistic world with its data, the tools and applications an agent uses, and tasks with verifiers. On Jul 9, 2026 it announced it would acquire Deeptune, a company that builds RL environments.
- Off-the-shelf datasets that are pre-built and ready to license, and custom projects that can cover agent environments and new verticals.
- The APEX benchmarks (the AI Productivity Index), which measure how models perform economically valuable tasks, plus leaderboards it hosts for partner benchmarks such as Harvey’s open-source legal benchmark, Harvey LAB.
- Evaluations for enterprises, and enterprise AI deployment with forward deployed engineers.
Considerations
- Its trust center lists SOC 2 Type 2 as of Nov 11, 2026, and no other certification.
- Its environment work is described across its research page, its dataset catalog and its blog, not on one product page.
Best for: Teams that want working professionals to write tasks, rubrics and evaluations, with benchmarks built on professional work.
Sources: mercor.com/mission, mercor.com, mercor.com/research, mercor.com/blog/mercor-to-acquire-deeptune, mercor.com/apex/off-the-shelf-data, mercor.com/apex, mercor.com/enterprise-evals, trust.mercor.com
Scale AI
Scale AI’s mission is “to develop reliable AI systems for the world’s most important decisions.” It provides training data, evaluations and AI applications to model developers, enterprises and governments. It was founded in 2016 and is headquartered in San Francisco.
Its founder, Alexandr Wang, is chairman of the board, and Francis deSouza has been CEO since Aug 10, 2026. On Jun 12, 2025 Scale AI announced an investment from Meta, said Meta would hold a minority of its equity, and said it remains independent.
What Scale AI offers
- Training data, annotations and RLHF through its Data Engine, with labels from domain experts.
- RL environments, introduced Feb 27, 2026 as Scale RL Environments and sold as Enterprise Atlas as of Nov 11, 2026, with MCP servers, CLI and GUIs across enterprise systems and desktop environments and expert-designed tasks and rubrics. Its environments run on AgentEnv, which it open-sourced on Oct 5, 2026.
- An off-the-shelf catalog of training data, evals and RL environments to license.
- Model evaluations and red teaming, plus research and public leaderboards from Scale Labs.
- AI applications for enterprises and governments, Scale GenAI Portfolio and Donovan.
- Contributors who work through Outlier, “a platform operated by Scale AI.”
Considerations
- Its RL environments are available by requesting samples.
- Its environments page covers tool use and computer use, so teams in other domains should ask what is available.
Best for: Teams that want data, evaluations and environments from one vendor, or that need a FedRAMP High authorized vendor (ask which products it covers).
Sources: scale.com/about, scale.com/blog/scale-appoints-new-ceo, scale.com/blog/scale-ai-announces-next-phase-of-company-evolution, scale.com/data-engine, scale.com/rlenvironments, scale.com/blog/rl-environments, labs.scale.com/blog/introducing-agentenv, scale.com/off-the-shelf-data, scale.com/experts, outlier.ai, scale.com/security
Key differences
- Age and breadth. Scale AI was founded in 2016 and also sells AI applications, Scale GenAI Portfolio for enterprises and Donovan for government. Mercor was founded in 2023 and sells enterprise evaluations and AI deployment next to its data work.
- Who does the work. Mercor describes a network of professionals vetted with AI interviews. Scale AI works with subject matter experts through Outlier, a platform it operates.
- RL environments. Scale AI introduced Scale RL Environments on Feb 27, 2026, sold as Enterprise Atlas as of Nov 11, 2026, and open-sourced AgentEnv, the framework its environments run on, on Oct 5, 2026. Mercor builds environments from worlds, tools, tasks and verifiers, and announced on Jul 9, 2026 that it would acquire Deeptune, an RL environment company.
- Benchmarks. Mercor’s APEX benchmarks measure professional tasks in fields such as law, banking, consulting and accounting. Scale AI publishes research and leaderboards through Scale Labs.
- Security. Scale AI lists SOC 2 Type II, ISO 27001, FedRAMP High and a DoD IL4 provisional authorization. Mercor lists SOC 2 Type 2.
How to choose
Choose Mercor if your tasks need working professionals, such as lawyers, bankers or physicians, to write and grade them, or if you want benchmarks built on professional work.
Choose Scale AI if you want data, evaluations and environments from one vendor, need government authorizations such as FedRAMP High, or also want AI applications for your own organization.
Either way, ask for samples, ask how graders are checked against reward hacking, and confirm the certifications your data requires.
Where Patronus AI fits
Patronus AI, which publishes this page, is a behavioral research lab for AI that builds RL environments and the data inside them, in coding, knowledge work, computer use and more. It sells a set of ready-made environments, though most of its work is custom, and domain experts review every environment before delivery.
As of Nov 11, 2026, from each company’s own pages.
| Criterion | Patronus AI | Mercor | Scale AI |
|---|---|---|---|
| What it is | A behavioral research lab that builds RL environments and data | An AI data company built on an expert network | A data, evaluation and applied AI company that also sells RL environments |
| What it sells | Ready-made and custom RL environments, and parts such as simulated apps and mock websites | Expert human data, evaluations, benchmarks and RL environments, plus custom AI agents for enterprises | RL environments on MCP servers, CLI and GUIs, next to human data, evaluations and applied AI for enterprises and governments |
| Quality checks | Domain expert review and automated checks on every environment, plus its own training runs | Rigorous tasks and verifiers for its RL environments, and expert-authored rubrics with an LM judge on its APEX benchmarks | Expert-designed tasks and rubrics, process and outcome verifiers, a rubric-based agent judge, and training runs on open models |
| Research | 20 papers as of Nov 11, 2026, including at NeurIPS, ICML and ACL | The APEX benchmarks, with papers on arXiv | Research papers on Scale Labs, and public leaderboards |
For more options, see Mercor alternatives, Scale AI alternatives and RL environment companies.
Frequently asked questions
What is the difference between Mercor and Scale AI?
Both supply expert human data, evaluations and RL environments to AI labs. Mercor, founded in 2023, is built around a network of domain experts and its APEX benchmarks. Scale AI, founded in 2016, also sells AI applications to enterprises and governments, and lists FedRAMP High.
Do Mercor and Scale AI sell RL environments?
Yes, as of November 2026. Scale AI sells them as Enterprise Atlas, part of Scale RL Environments, which it introduced Feb 27, 2026. Mercor builds them from worlds, tools, tasks and verifiers, and announced on Jul 9, 2026 that it would acquire Deeptune, an RL environment company.
Which security certifications do Mercor and Scale AI list?
As of Nov 11, 2026, Mercor’s trust center lists SOC 2 Type 2. Scale AI’s security page lists SOC 2 Type II, ISO 27001, FedRAMP High and a DoD IL4 provisional authorization.
What are the alternatives to Mercor and Scale AI?
Other companies that supply expert data or RL environments include Surge AI, Turing, Snorkel AI, Labelbox, Invisible and Patronus AI. Our Mercor alternatives and Scale AI alternatives pages compare them.
Is there an alternative to both?
Yes. Mercor and Scale AI both sell expert data alongside RL environments. For teams whose main need is the environments themselves, Patronus AI, which publishes this page, is a behavioral research lab that builds ready-made and custom RL environments and the data inside them. Talk to Patronus AI, or see the other options on the alternatives pages linked above.
How was this comparison made?
Patronus AI wrote it from each company’s own pages, read on Nov 11, 2026 and linked under each company’s section. It is a dated snapshot. Products, leaders and certifications change, so check each company’s site before you buy.
Snapshot as of