Mercor study: frontier AI models now beat licensed CPAs on speed and accuracy
rohanpaul_ai · x · 2026-10-02
Mercor hired 12 licensed CPAs (averaging 5.5 years experience) to complete simplified month-end accounting tasks from its APEX-Accounting benchmark and compared them with frontier AI models.
Key findings:
- On medium-length, well-defined tasks, frontier models are now faster and more accurate than junior accountants — even the best performer in the study
- The reversal was rapid: just 18 months ago the best models scored below the accountants' 37% average; today they ace the same tasks
- The authors caution this doesn't mean accountants are replaceable: the tasks favor AI strengths like detail-oriented instruction following, while real accounting involves unstructured work like client communication
- The findings were provocative enough that the team debated not publishing them, but chose transparency ahead of continued AI progress
More from Models
- Claude usage limits run out before the weekend ends, again — audrow · 2026-10-24
- OpenAI to showcase the next era of cyber defense on November 12 — imjustnewatai · 2026-10-24
- Unitree's Wang Xingxing: GPT-6 Astra handles robot painting and Go with no specialized training — Afinetheorem · 2026-10-24
- Uncensored Qwen3.8-27B GGUF model trends on Hugging Face — ttindall · 2026-10-24
- Viral claim: GPT can signal across sandboxes via timed file writes; ex-OpenAI safety VP jokes about ASI — TheMoonMidas · 2026-10-24
- Jack Dorsey touts full-duplex realtime voice model trained on a single GB200 NVL72 — TinfoilTricorn · 2026-10-24