Oneeno
Access

ONEENO / EDUCATION & TRAINING

Smart is the beginning.
Staying current is the difference.

We do not promise you titles — you cannot verify them. What we do instead: put on the table the real exams the engine behind your team has been measured on, train your team on your practice, and keep it learning every quarter. This is how that works.

THE BENCHMARK / WHAT THE ABBREVIATIONS MEAN

Real exams.
Published results.

The benchmark line on every card of your business team refers to these exams. They apply to the model class your team runs on — with sources at the bottom of this page.

YOUR CFO AND YOUR TAX SPECIALIST

CFA I·II·III

the three-part exam for financial analysts.

In an independent study, the model class behind your CFO passes all three levels; level I with 97.6%. People take four years on average, and fewer than half pass each level.

How it was measured

Researchers from Columbia University, Rensselaer and UNC gave six reasoning models 980 official practice exam questions: three full level I exams (540 multiple-choice questions), level II case questions and level III including open answers — scored against the usual CFA Institute pass mark.

Source: Reasoning Models Ace the CFA Exams (arXiv, December 2025) ↗

YOUR LAWYER AND YOUR TAX SPECIALIST

UBE (bar exam)

the American admission exam for lawyers.

Passed, with a reported score in the 90th percentile — better than nine out of ten human candidates in that measurement.

How it was measured

Katz and colleagues had the model sit the full exam — multiple choice (MBE), essays and performance tests — graded to the official standard. A later re-analysis arrived at a lower percentile, but still well above the pass mark of every US state that uses this exam.

Source: GPT-4 Passes the Bar Exam (Katz et al., 2023; commentary Stanford Law) ↗

YOUR PROFESSOR AND YOUR RESEARCHER

GPQA

PhD-level science questions, deliberately "Google-proof".

PhD holders score 65% in their own field. The newest models score around 91% — across all fields at once.

How it was measured

GPQA was written by PhD students in physics, chemistry and biology, and designed so that looking things up does not help: skilled non-experts with internet access score only 34%. PhD experts reach 65% within their own field; the newest models’ scores are on a public leaderboard that is kept up to date.

Source: GPQA benchmark (Rein et al.) and the Epoch AI leaderboard ↗

YOUR ADMINISTRATOR (PERSONAL TEAM)

CPA

the American accountancy exam.

Passed by the same generation of models; the foundation under the number work of your administration.

How it was measured

Exam trainer Kaplan analysed how language models perform on the parts of the CPA exam (accounting, audit, tax law and business environment) and saw the new generation pass the mark — an exam human candidates study for more than a year on average alongside their job.

Source: Kaplan, Generative AI impacts credentials ↗

Honesty first: these are results of the underlying model class in published studies, not of an individual conversation. Your team carries no protected titles, is not an accountant or attorney of record, and does not replace legally required advisers. You judge the outcome — and you can, because every analysis shows its sources and assumptions.

THE METHOD / HOW DO YOU COMPARE AN AI WITH A DIPLOMA?

No diploma.
But the same exam.

An honest question deserves an honest answer: this is how such a comparison works — and how it does not.

01 / THE REAL EXAM

The same questions as human candidates.

Independent researchers give the models the official practice and sample exams from the real exam bodies — the same questions people use to prepare. No home-made quiz, no selection of easy questions.

02 / THE SAME YARDSTICK

Scored against the official pass mark.

Answers are graded to the exam’s own standard: the CFA Institute pass mark, the bar exam percentiles, the GPQA expert score. So “passed” means the same here as for a human candidate.

03 / PUBLISHED AND VERIFIABLE

Every claim has a study.

All results on this page come from published research with authors and dates — expandable per exam above, with the source attached. We add nothing and round nothing up; you can read it all without leaving our site, and then at the source itself.

04 / WHAT IT DOES AND DOES NOT SAY

Knowledge is measured. Authority is not.

An exam tests knowledge and reasoning; a diploma also grants a legal authority — which an AI does not have. The results apply to the model class (the engine behind your team), not to a certificate, and conditions differ. That is why we say “passes the exam in a published study” and never “holds the diploma”. For work that requires legal authority, your team works towards your accountant, lawyer or tax adviser.

THE TRAINING / THREE TIERS ON YOUR PRACTICE

Not just educated.
Trained on you.

An exam says what the engine can do. The tiers say what your team knows about you — and that grows.

Tier 1 — Inducted

Knows your profile, your professional language and your projects. Every team member starts here, in the first days after the start. Not one generic answer, but your context as the starting point.

Tier 2 — Trained on your practice

Works with your real material: the CFO with your numbers and budget, the Lawyer with your contracts, the minute-taker with your meetings. Every specialist delivers trial work that you judge — before you build on it.

Tier 3 — Kept sharp

Back to school every quarter: new sources in the knowledge table, your corrections built into the way of working, the results measured again. You receive a short report of what got better.

ALWAYS UP TO DATE

A staff that never goes out of date.
And never misses a day.

Read up every night

Your scouts read the sources that belong to your world every night. What changed overnight, your team knows by breakfast.

Every correction a lesson

If you say “that is not what I mean”, it is not a complaint but course material. Your preferences and your language become part of how your team works.

Along with every model generation

Your team runs on the newest model class and moves up as soon as a better one arrives. A human staff has to be recruited and trained again and again; this one gets smarter every year — without a hiring round.

Accountability every quarter

The continued training is not a promise but a report: what was learned, what improved, where the level stands now.

Sources: Reasoning Models Ace the CFA Exams (arXiv, dec 2025) · CNBC (sept 2025) · GPT-4 Passes the Bar Exam (Stanford Law, 2023) · GPQA: A Graduate-Level Google-Proof Q&A Benchmark · GPQA Diamond leaderboard · Kaplan (CPA) · salary sources for the value indication: CMweb (CFO) · DRB Groep.

Meet your two teams
THE NEXT STEP

Discover your edition.