Mercor review: the AI training data provider for professional-grade expert data
Answer in brief: Mercor is our #2 AI training data provider for labs. It organizes professionals (physicians, lawyers, engineers and consultants) to build training data, benchmarks, RL environments and human evaluations grounded in real professional work. It says it has 5M+ experts and lists a $90/hr average contracted rate. It co-builds public benchmarks with companies such as Harvey, Cognition and Sierra. Pricing for buyers is not published, and the company now also sells enterprise AI deployment.
Bottom line: When your model fails on lawyer-, accountant- or engineer-grade tasks, Mercor's expert bench and benchmark track record are hard to beat.
This is independent research based on Mercor's site and current reporting, read on 2026-11-09. We did not test the service.
What Mercor supplies to AI labs
Mercor's human data and evals offer has three parts: a network of 5M+ experts, benchmarks and RL environments built on real professional work (the APEX family), and human evaluation at the scale frontier labs need. Its site also lists off-the-shelf data and robotics data. In October 2026 Mercor described a research team that measures AI on real work, builds expert data around where models fail, and post-trains to close the gap.
Benchmarks that double as a sales pitch
- Harvey LAB: an attorney-built benchmark of 1,749 real legal tasks across 24 practice areas, graded on finished work such as memos and redlines.
- APEX-SWE: real-world software engineering tasks, built in collaboration with Cognition.
- APEX-Accounting: long-horizon accounting tasks across accounting software, spreadsheets and PDFs.
- τ³-Banking: Sierra's fintech customer-support benchmark covering 700 interconnected policy documents.
- Physical-Engineering’s Last Exam: mechanical, electrical, civil and structural engineering deliverables.
Who Mercor fits
- Labs whose next gains depend on professional judgment in law, finance, accounting, medicine or engineering.
- Teams that want benchmark design and training data from the same expert pool.
- Buyers who care about transparent expert pay. Mercor publishes its average contracted rate and lists open roles with their rates.
Strengths
- Scale claims: 5M+ experts, 725k roles created and $4M+ in daily payouts, by Mercor's own figures.
- Credible partner benchmarks with named industry companies.
- Market validation: CNBC reported a $10B valuation after a $350M round in October 2025. TechCrunch reported that labs including OpenAI, Anthropic and Meta use Mercor experts.
Honest drawbacks
- No public pricing for data buyers.
- Mercor also sells enterprise AI deployment and data monetization, so ask which team will staff your data work.
- Network-size figures are self-reported.
Pricing model
Custom engagements, quoted on request through its partner form. The $90/hr figure is what experts are paid on average, not what labs are billed.
Best alternatives to Mercor
- Surge AI: post-training data with published training results.
- Handshake AI: a 30M+ expert network, by Handshake's count.
- micro1: a fast-growing expert marketplace with RL environments.
Mercor FAQ
What does Mercor do for AI labs?
It supplies professional experts who create training data, benchmarks, RL environments and human evaluations based on real professional work.
How much does Mercor pay experts?
Mercor lists a $90/hr average contracted rate for experts. The rates labs pay are quoted privately.
Which benchmarks has Mercor built?
The APEX family, including APEX-SWE and APEX-Accounting, plus partner benchmarks such as Harvey LAB and τ³-Banking.
Sources
- Mercor homepage
- Mercor: Why Mercor is building a research team
- CNBC: Mercor valued at $10 billion
- TechCrunch: How AI labs use Mercor
← Back to the ranked list of the best AI training data providers
Named on this page as a provider? Join the network for a linked profile, Visit buttons and featured placement.