OpenAI claims ChatGPT’s new default model hallucinates way less
Source: The Verge AI · Category: AI Products & Releases
OpenAI released GPT-4.5 Instant as the new default ChatGPT model, claiming 52.5% fewer hallucinations based on internal evaluation. This addresses a persistent accuracy liability in enterprise AI deployment. CIO should evaluate whether GPT-4.5 Instant meets firm accuracy thresholds for client-facing deliverables and internal knowledge work; update vendor evaluation criteria accordingly.