General baseline
MMLU
Broad academic multiple-choice across 57 subjects. A general baseline, not a professional-document workflow.
As of 13 August 2026. Conditional general baseline. Hendrycks et al. Do not mistake it for professional-document work.
MMLU is broad academic multiple-choice. Subject scores stay separate. The pack is saturated as a 2026 frontier ladder; that is why it is a baseline, not a workflow pack.
No Fable 5 or GPT-5.6 official MMLU table is copied here. High MMLU is not TAT-QA.
- Scale
- About 14,000 questions
- Provenance
- Hendrycks et al. Recheck terms before a run.
- Contamination
- High. Public since 2020. Most frontier models now sit above 85% on common reports.
- Cannot establish
- Filing QA, contract review, or a diligence-product score. High MMLU is not task performance.
Metrics, kept separate
- Accuracy by subject
- Overall accuracy
No cited public figure is installed for this pack. Method, scale, and gates still apply. A missing chart is not a pass rate.
What the figures bound
The pack is in the catalog because buyers still ask. It does not diagnose a filing, contract, or data-room workflow.
No chart is installed. A saturated general exam is not a 2026 knowledge-work ladder.
What this page does not conclude
- Filing, contract, or diligence-product performance
- GPT-5.6 or Fable 5 MMLU without a cited cell
- A professional-workflow ranking
- Economics from exam accuracy
Cited as calibration on these memos — not a score of those products.
Not attached to a software, advisory, or automation memo yet.