AI Model Outperformed Certified Accountants in Benchmarks

A Mercor study shows AI models can now complete month-end accounting tasks faster and more accurately than human professionals.

Updated on Oct. 4, 2026 in Artificial Intelligence

AI Model Outperformed Certified Accountants in Benchmarks

Live Poll

Do you worry that AI will eventually make your current job role obsolete?

A recent study by Mercor found that the Claude Opus 5 model achieved 100 percent accuracy on month-end accounting benchmarks, while 12 certified accountants averaged 37 percent. This research-stage comparison highlights a significant performance gap in structured financial tasks.

Why it matters

The findings establish a clear human baseline for AI performance in accounting, suggesting that highly structured workflows are increasingly susceptible to automation. This shift could significantly enhance productivity in the accounting sector as software adoption scales.

Claude Opus 5 demonstrated a 100 percent accuracy rate on 20 test attempts, compared to the 69 percent accuracy achieved by top-tier AI models 18 months prior. The model completed tasks in under 10 minutes, significantly faster than the 30 to 180 minutes required by human accountants.

The players

Mercor

A research entity that conducts benchmarking studies comparing artificial intelligence capabilities against human professional standards.

Claude Opus 5

An artificial intelligence model developed to handle complex logical and structured reasoning tasks.

The details

Researchers evaluated performance by testing both human accountants and AI on identifying document details, calculating figures, and tabulating results across four month-end scenarios. These tests utilized real company records to measure efficiency and precision. Cost comparisons were derived by benchmarking the operational expense of the AI model against average American accountant salary data.

April 2025: Previous top-tier models reached 69 percent accuracy.
October 4, 2026: Mercor study results were published.

The Tech Race

This study underscores a rapid trajectory in machine performance, showing a 31 percentage point gain in accuracy for top-tier models since April 2025. It positions AI systems as increasingly viable tools for tasks that historically required years of human professional experience.

Accounting professionals should expect AI to begin taking over highly structured, repetitive month-end workflows. While these tools currently offer significant cost and speed advantages, their real-world deployment will depend on individual firm integration and verification standards.

The takeaway

The gap between human speed and AI efficiency in structured accounting is widening, as demonstrated by the cost and time savings identified in this benchmark. Readers should track upcoming industry productivity reports to see how firms incorporate these findings into their operations.