OpenAI has introduced its latest default ChatGPT model, GPT-5.5 Instant, which shows marked improvement in reducing hallucinations—errors where the AI fabricates information. Internal evaluations reveal that this new model generates 52.5% fewer hallucinated claims compared to the previous Instant model (GPT-5.3), particularly in critical fields like medicine, law, and finance. Additionally, GPT-5.5 Instant cuts inaccurate claims by 37.3% in conversations previously flagged for errors, reflecting OpenAI’s ongoing commitment to enhancing factual accuracy in AI models.
Back
