New OpenAI models hallucinate more often than their predecessors

OpenAI's latest reasoning models, o3 and o4-mini, have demonstrated a worrying increase in hallucination rates, according to the PersonQA evaluation. With hallucination rates of 33% for o3 and 48% for o4-mini, these models actually perform worse than their predecessors like o1, which had a rate of 16%. Despite being designed for improved reasoning using more computing power and refined strategies, these models appear to have taken a step back in their ability to provide accurate information, highlighting ongoing challenges in AI development that require further investigation.

OpenAI's latest reasoning models, o3 and o4-mini, exhibit significantly higher hallucination rates compared to previous models, signaling challenges for AI accuracy.

The PersonQA evaluation reveals that o3 has a hallucination rate of 33 percent, while o4-mini reaches as high as 48 percent, nearly half the time.

Read at Techzine Global