The biggest problem in AI right now is not capability. It is measurement.
No reliable method exists to assess whether an AI model can perform valuable professional work in law, finance, accounting, scientific research or consulting. Academic benchmarks measure what is easy to test. They do not measure what matters in practice.
Mercor aims to fix that. The company is paying researchers up to $80000 to help.
The Mercor Research Fellowship APEX is now open. Applications are accepted on a rolling basis. The current deadline is 26 September 2026. Earlier applications receive more attention.
At a Glance
| Detail | Information |
|---|---|
| Fellowship | Mercor Research Fellowship APEX |
| Organization | Mercor |
| Duration | 3 months or 6 months |
| Funding | $40000 for 3 months or $80000 for 6 months |
| Commitment | Minimum 30 hours per week |
| Location | Remote or San Francisco London New York |
| Who can apply | International Applicants |
| Application deadline | 26 September 2026 rolling |
What Mercor Is Trying to Do
Mercor runs the APEX benchmark family. These evaluations test whether AI models can perform real professional work rather than standardized academic tasks.
The company works with domain experts worldwide to train and evaluate frontier AI systems. Its research examines how AI is changing labor markets, professional workflows and firm structures. It uses proprietary datasets, enterprise partnerships and large expert networks.
The fellowship brings external researchers into that work. Selected fellows develop new benchmarks, evaluation methods or economic studies that fill gaps in current AI measurement.
Fellows do not assist existing projects. The research proposal comes from the applicant.
The Funding Package
The financial support is substantial.
- 3 month fellowship: $40000 stipend
- 6 month fellowship: $80000 stipend
Fellows also receive:
- Unlimited API credits for frontier AI models
- Dedicated GPU compute resources
- Budget for expert reviewers and human data collection
- Weekly one on one mentorship with the APEX research team
- Access to Mercor’s evaluation infrastructure
- Connections with researchers across academia and leading AI labs
- Consideration for a full time position with the APEX team
Strong fellowship performance can lead to a permanent role. The funding the infrastructure and a potential career path sit in the same place.
Who Should Apply
The fellowship is open to a wide range of backgrounds. Mercor is not looking for one type of researcher.
Strong applicants often come from:
- Economics
- Computer science or machine learning
- Statistics or data science
- Social sciences or human computer interaction
- Psychometrics
- Law, finance, medicine or scientific research
What matters most is the quality of the proposal. Fellows are selected for a clearly defined original research idea. Credentials alone do not decide the outcome.
Practical requirements include:
- Availability for at least 30 hours per week
- Ability to commit for three to six months
- Capacity to work independently in a fast moving environment
Graduate students are welcome alongside working researchers and professionals.
What the Research Focuses On
Mercor lists priority areas. Proposals outside these themes are also considered.
Current priority areas include:
- Long-horizon AI agent evaluations in law, finance and consulting
- Advanced software engineering benchmarks
- Accounting and financial workflow evaluations
- AI in scientific research including mathematics, biology, materials science and physics
- Economic benchmarking of negotiation management and decision-making
- Evaluation methodology research such as contamination resistance and human versus model assessment
- Economics of human data marketplaces and AI labor platforms
- AI native organizational design and the future of firms
- Novel benchmark designs and innovative evaluation frameworks
The core question is how to accurately measure what AI can do in the real world. Researchers with original answers from any direction are encouraged to apply.
What Fellows Actually Do
Day to day work varies by project. It typically includes:
- Designing benchmark tasks and evaluation frameworks
- Developing grading systems and scoring methods
- Validating benchmark quality with subject matter experts
- Conducting empirical economic research using proprietary datasets
- Running experiments with frontier AI models
- Publishing findings through papers, datasets, leaderboards or research reports
Fellows work closely with Mercor’s APEX team. Mentorship is weekly and one on one.
The Proposal Is the Centerpiece
The application centers on a research proposal. This is not a supporting document. It is the main element.
Mercor selects based on ideas. A researcher with a shorter publication list but a sharp, specific proposal for a benchmark or evaluation method that fills a real gap will outperform a stronger CV with a vague direction.
Before writing the proposal, ask three questions:
- What specific gap in AI evaluation or economic research am I addressing
- Why has this not been done well yet
- What would a successful outcome look like and how would it change how the industry measures AI
Clear answers to these questions form the backbone of a competitive application.
Remote or In Person
The fellowship supports both options.
Remote participation is fully available. Fellows who prefer in person work can join Mercor offices in San Francisco, London or New York City.
For international applicants, the remote option makes the fellowship accessible from any location. Check visa requirements if you plan in person participation in the US or UK.
Why Applying Early Matters
The 26 September 2026 deadline is the current confirmed cutoff. Rolling review changes the practical calculation.
Applications reviewed earlier compete against a smaller pool. They receive more careful attention from the APEX team before deadline pressure builds.
If your proposal is ready or close to ready submit it now rather than refining language for weeks.