Claim: Grok 4.5's hallucination rate jumped from 25% to 54% compared to Grok 4.3

The claim that Grok 4.5's hallucination rate jumped from 25% to 54% compared to Grok 4.3 is mostly true. Independent sources, including benchmarks from AI testing platforms, support this increase. However, some alternative sources dispute this finding, suggesting that Grok 4.5's hallucination rate is lower than Grok 4.3, which raises questions about the consistency of the data reported across different evaluations. The models diverge sharply — treat this as higher-uncertainty. Gemini comes in highest (85%), while Perplexity is lowest (50%). Gemini expresses higher confidence than Perplexity on this claim. While the majority of sources indicate that Grok 4.5's hallucination rate increased significantly, the conflicting report from an alternative source suggests that the rate is actually lower than Grok 4.3. This discrepancy may arise from differences in testing methodologies or benchmarks used. As a result, while the evidence leans towards the claim being true, the existence of contradictory data introduces a level of uncertainty regarding the absolute accuracy of the reported figures.