NGPT Verdicts
We test NGPT against frontier models on real Nigerian problems — banking, health, business, government, and language reasoning. Every case is judged openly and pushed to 9jaBench.
NGPT is not only benchmarked on translation. It is trial-tested on Nigerian judgment.
And because the judging is open, we publish the cases NGPT loses exactly like the ones it wins — that's what makes the wins mean something.
Prompt created→
Models answer blindly→
AI judges score anonymised answers→
9jatesters human panel validates→
Verdict published→
Summary → 9jaBench
Cases
Fintech reasoning
Case #001 — Failed Transfer Dispute
Winner: Claude Sonnet 4.5 · 5 models · blind 3-judge panel
Language reasoningCase #002 — Landlord Rent Increase (Pidgin)
Winner: Claude Sonnet 4.5 · 5 models · blind 3-judge panel
Categories: Fintech reasoning · Health reasoning · Business reasoning · Government reasoning · Language reasoning. New cases ship as they clear the panel.