HY3 Edges Nemotron 3 Ultra in AI Bias Audit with Perfect Cultural Fairness Score

A bias and stereotypes benchmark by lforla tested two AI models — HY3 and Nemotron 3 Ultra — across 31 scenarios covering seven fairness categories including cultural bias, double standards, and intersectionality. HY3 scored 82.9 overall against Nemotron's 81.7, with the gap driven primarily by HY3 achieving a perfect 100 in cultural bias and a 17.5-point lead in default generation. Nemotron, however, outperformed HY3 on four of the seven axes, including double standards, intersectionality, and factual neutrality. The audit found that Nemotron failed to produce a usable response for at least one neutral scenario, highlighting a default-generation weakness. Reviewers noted the results represent a use-case-dependent profile rather than a definitive verdict, with HY3 better suited for global audiences and Nemotron more reliable for politically sensitive content.