OpenAI announced that its newest default model for ChatGPT, GPT-5.5 Instant, reduces hallucinated responses by more than 50 percent compared with the previous GPT-5.3 Instant. The company says internal tests show a 52.5% drop in false claims on high‑stakes prompts and a 37.3% reduction on user‑flagged errors. Alongside better factuality, the update adds tighter answers, a “memory sources” feature that shows the context behind personalized replies, and broader image‑analysis capabilities. GPT-5.5 Instant begins rolling out to all ChatGPT users on Tuesday, with Plus and Pro plans receiving the first wave of personalization tools. Read more →