Mercor

Mercor: human data and evaluation for artificial intelligence development

Mercor presents itself as a platform that organizes human intelligence to power the AI economy. Its activity focuses on connecting expert professionals with cutting-edge artificial intelligence labs and companies that need data, evaluations, and training environments for their models. The company operates from San Francisco, California.

Network of experts and jobs for professionals

The platform has a network of more than 30,000 experts, including physicians, lawyers, engineers, and consultants. Through it, professionals can find paid work solving AI challenges. According to published data, the average contracted rate is $109 per hour, 466,000 roles have been created, and daily payouts have exceeded $4 million.

The site shows examples of active roles, such as surveys for clinicians on ambulatory and community care, a general clinician expert pool (MD/DO) at a rate of $150 per hour, and an opinion study for Devoted Health Medicare Advantage members at a rate of $120 per hour. These data correspond to the catalog visible at the time of analysis.

APEX Benchmarks: evaluating AI models

Mercor develops the APEX family of benchmarks, which assesses whether frontier AI models can perform economically valuable tasks in professional services, medicine, and software engineering. The family includes:

• APEX-Agents: long-horizon, cross-application tasks in professional services. The results shown include Opus 5.5 Max at 73.5%, Fable 5.1 Max at 68.6%, and Gemini 3.7 Flash High at 67.8%.
• APEX-Accounting: long-horizon, cross-application tasks in professional accounting, in collaboration with Ramp. The results include Opus 5.5 Max at 61.8%, Fable 5.1 Max at 61.0%, and Opus 5.5 Medium at 59.9%.
• APEX-SWE: real-world software engineering across integration and observability, in collaboration with Cognition. The results include Opus 5 Max at 63.7%, Fable 5.1 Max at 63.6%, and Fable 5 Max at 58.8%.

Enterprise solutions

Mercor offers enterprises the ability to turn their expertise into production AI, new revenue streams, and the human intelligence behind leading models. Its services are organized into three areas:

• Human data & evals: access to more than 30,000 experts (physicians, lawyers, engineers, consultants), benchmarks and reinforcement learning environments built on real professional work (APEX), and human evaluation at the scale frontier labs rely on.
• Data monetization: creation of new revenue streams from the workflow data that enterprises already generate. The platform applies automated anonymization that masks more than 60 categories of sensitive identifiers, and the organization retains full control over what data is shared.
• Enterprise AI: identification of high-value opportunities by mapping workflows and institutional knowledge, deployment of production AI with embedded Forward Deployed Engineers and enterprise-grade integrations, and measurement and improvement of AI performance with a platform for context, evaluations, and continuous optimization.

Acquisition of Deeptune and valuation

The site announces that Mercor will acquire Deeptune, a transaction that, according to the company, will make it possible to build realistic training environments across more industries and roles by combining Mercor's five million experts with Deeptune's platform. It also includes a CNBC news item according to which Mercor reached a valuation of $10 billion after a $350 million funding round. Other media mentions include Forbes, TIME, and Bloomberg.

Editorial activity and content

The company maintains a blog and a news section with articles on enterprise AI, AI agents, and their application in sectors such as insurance. Among the published titles are analyses of why AI agents get stuck in proof of concept and how AI should earn underwriting authority.