OpenAI claims ChatGPT’s new default model hallucinates way less

OpenAI's newest default exemplary for ChatGPT mightiness not marque worldly up arsenic much. Hallucinations person been an ongoing problem for AI models, but OpenAI says its caller GPT-5.5 Instant exemplary has "significant improvements successful factuality crossed the board."

The institution claims that, based connected "internal evaluations," GPT-5.5 Instant produced "52.5% less hallucinated claims" than its Instant exemplary for GPT-5.3 "on high-stakes prompts covering areas similar medicine, law, and finance." GPT-5.5 Instant besides "reduced inaccurate claims by 37.3% connected particularly challenging conversations users had flagged for factual errors."

OpenAI besides claims that GPT-5.5 Instant is …

Read the afloat communicative astatine The Verge.