The Mercor Research Fellowship — APEX is offering researchers and domain experts a chance to build new AI benchmarks and study the economics of artificial intelligence, with awards of $40,000 for a three-month fellowship or $80,000 for a six-month fellowship. The program is open remotely, can also be done in person from Mercor’s San Francisco office where appropriate, and is aimed at projects that evaluate whether AI systems can perform economically valuable professional work.
Key facts
- The fellowship funds proposal-based research rather than a predefined project.
- Selected fellows can receive access to frontier AI models, GPU computing resources, expert human-data support, proprietary datasets, and mentorship.
- The application deadline is September 26, 2026, at 9:00 AM GMT+3.
- Applicants are expected to pitch a benchmark, evaluation methodology, or economic study they want to develop.
- Mercor says fellows should commit at least 30 hours per week, and full-time participation is preferred.
What is Mercor APEX trying to measure?
Mercor says APEX is designed to test whether frontier AI systems can do economically valuable work in realistic professional settings, not just answer short questions or solve narrow coding puzzles. The company’s focus is on long-horizon, multi-step tasks and expert-designed evaluations that better reflect real-world work.
The benchmark family is meant to examine tasks such as investment banking workflows, corporate law assignments, accounting, software engineering, financial workflows, and research-level reasoning. Mercor says these benchmarks are developed and validated with domain experts including lawyers, accountants, software engineers, scientists, consultants, and finance professionals.
What this means for researchers and applicants
For researchers, PhD candidates, independent scholars, and domain experts, the fellowship is structured as an opportunity to turn a specific idea into a completed project. Mercor says fellows can design, implement, validate, and potentially publish work on new benchmarks, evaluation methods, or economic analyses using its infrastructure and data access.
The program may be especially relevant to applicants who want to study AI evaluation, AI agents, or how artificial intelligence affects labor markets, professional services, productivity, and organizational workflows.
How does the fellowship work?
Mercor describes the fellowship as rolling admissions, with options for three-month or six-month terms. It supports remote participation and may allow in-person work from San Francisco. Applicants are expected to submit a concrete project proposal, and Mercor says the quality, originality, relevance, and feasibility of that pitch are likely to matter in selection.
Selected fellows may work on benchmark design, grading rubrics, task piloting, scoring systems, or economic studies using Mercor’s proprietary datasets and internal research resources.
What kinds of projects are Mercor looking for?
Mercor highlights several areas of interest, including long-horizon agentic tasks in law, finance, and consulting; software engineering evaluation; professional accounting and finance workflows; AI for science; and new economic benchmarks. The company also says strong proposals outside those categories are welcome.
Possible outputs include academic papers, research reports, open datasets, new AI benchmarks, public leaderboards, evaluation methodologies, and internal research frameworks.
Frequently asked questions
Who can apply for the Mercor Research Fellowship APEX?
Mercor says the fellowship is intended for researchers and domain experts with a specific research proposal. It may suit people taking academic leave, doing summer research, working through a flexible stage of a PhD, transitioning between roles, conducting independent research, or exploring a career in AI evaluation.
What is the application deadline?
The application deadline is September 26, 2026, at 9:00 AM GMT+3.
